Klasifikasi PEDANTO InceptionV3¶
Modules dan Packages¶
In [ ]:
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
import tensorflow.keras
from tensorflow.keras import backend as K
import tensorflow.keras.utils
from glob import glob
from tensorflow.keras.models import Sequential
from sklearn.model_selection import train_test_split
from tensorflow.keras.layers import Input, Lambda, Dense, Flatten, Conv2D, MaxPooling2D, Dropout, Activation, BatchNormalization
from tensorflow.keras.models import Model
from tensorflow.keras.models import load_model
from tensorflow.keras.applications.inception_v3 import InceptionV3
from keras.applications.vgg16 import VGG16
from tensorflow.keras.applications.inception_v3 import preprocess_input
from tensorflow.keras.preprocessing import image
from tensorflow.keras.preprocessing.image import ImageDataGenerator,load_img, array_to_img, img_to_array
%matplotlib inline
K.clear_session()
Import Dataset¶
In [ ]:
from google.colab import drive
drive.mount("/content/gdrive")
Drive already mounted at /content/gdrive; to attempt to forcibly remount, call drive.mount("/content/gdrive", force_remount=True).
In [ ]:
# Define Constants by re-sizing all the images
IMAGE_SIZE = [224, 224]
train_path = '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train'
valid_path = '/content/gdrive/My Drive/Colab Notebooks/dataset8020/val'
Data Preprocessing¶
In [ ]:
def normalize(input_image, input_mask):
input_image = tf.cast(input_image, tf.float32) / 255.0
input_mask -= 1
return input_image, input_mask
InceptionV3¶
In [ ]:
# Import the InceptionV3 model and here we will be using imagenet weights
inception = InceptionV3(input_shape=IMAGE_SIZE + [3], weights='imagenet', include_top=False)
In [ ]:
# We don't need to train existing weights
for layer in inception.layers:
layer.trainable = False
In [ ]:
# Folders in the Training Set
folders = glob('/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/*')
folders
Out[ ]:
['/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Spider_mites', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Septoria_leaf_spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Late_blight', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Mosaic_virus', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Leaf_Mold', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Healthy', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Bacterial_spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Early_blight', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Target_Spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Yellow_Leaf_Curl_Virus']
Membuat Arsitektur Model InceptionV3¶
In [ ]:
# Model layers -> can add more if required
x = Flatten()(inception.output)
prediction = Dense(len(folders), activation='softmax')(x)
In [ ]:
# Create a model object
model = Model(inputs=inception.input, outputs=prediction)
# View the structure of the model
model.summary()
Model: "model_1"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_2 (InputLayer) [(None, 224, 224, 3 0 []
)]
conv2d_94 (Conv2D) (None, 111, 111, 32 864 ['input_2[0][0]']
)
batch_normalization_94 (BatchN (None, 111, 111, 32 96 ['conv2d_94[0][0]']
ormalization) )
activation_94 (Activation) (None, 111, 111, 32 0 ['batch_normalization_94[0][0]']
)
conv2d_95 (Conv2D) (None, 109, 109, 32 9216 ['activation_94[0][0]']
)
batch_normalization_95 (BatchN (None, 109, 109, 32 96 ['conv2d_95[0][0]']
ormalization) )
activation_95 (Activation) (None, 109, 109, 32 0 ['batch_normalization_95[0][0]']
)
conv2d_96 (Conv2D) (None, 109, 109, 64 18432 ['activation_95[0][0]']
)
batch_normalization_96 (BatchN (None, 109, 109, 64 192 ['conv2d_96[0][0]']
ormalization) )
activation_96 (Activation) (None, 109, 109, 64 0 ['batch_normalization_96[0][0]']
)
max_pooling2d_4 (MaxPooling2D) (None, 54, 54, 64) 0 ['activation_96[0][0]']
conv2d_97 (Conv2D) (None, 54, 54, 80) 5120 ['max_pooling2d_4[0][0]']
batch_normalization_97 (BatchN (None, 54, 54, 80) 240 ['conv2d_97[0][0]']
ormalization)
activation_97 (Activation) (None, 54, 54, 80) 0 ['batch_normalization_97[0][0]']
conv2d_98 (Conv2D) (None, 52, 52, 192) 138240 ['activation_97[0][0]']
batch_normalization_98 (BatchN (None, 52, 52, 192) 576 ['conv2d_98[0][0]']
ormalization)
activation_98 (Activation) (None, 52, 52, 192) 0 ['batch_normalization_98[0][0]']
max_pooling2d_5 (MaxPooling2D) (None, 25, 25, 192) 0 ['activation_98[0][0]']
conv2d_102 (Conv2D) (None, 25, 25, 64) 12288 ['max_pooling2d_5[0][0]']
batch_normalization_102 (Batch (None, 25, 25, 64) 192 ['conv2d_102[0][0]']
Normalization)
activation_102 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_102[0][0]']
conv2d_100 (Conv2D) (None, 25, 25, 48) 9216 ['max_pooling2d_5[0][0]']
conv2d_103 (Conv2D) (None, 25, 25, 96) 55296 ['activation_102[0][0]']
batch_normalization_100 (Batch (None, 25, 25, 48) 144 ['conv2d_100[0][0]']
Normalization)
batch_normalization_103 (Batch (None, 25, 25, 96) 288 ['conv2d_103[0][0]']
Normalization)
activation_100 (Activation) (None, 25, 25, 48) 0 ['batch_normalization_100[0][0]']
activation_103 (Activation) (None, 25, 25, 96) 0 ['batch_normalization_103[0][0]']
average_pooling2d_9 (AveragePo (None, 25, 25, 192) 0 ['max_pooling2d_5[0][0]']
oling2D)
conv2d_99 (Conv2D) (None, 25, 25, 64) 12288 ['max_pooling2d_5[0][0]']
conv2d_101 (Conv2D) (None, 25, 25, 64) 76800 ['activation_100[0][0]']
conv2d_104 (Conv2D) (None, 25, 25, 96) 82944 ['activation_103[0][0]']
conv2d_105 (Conv2D) (None, 25, 25, 32) 6144 ['average_pooling2d_9[0][0]']
batch_normalization_99 (BatchN (None, 25, 25, 64) 192 ['conv2d_99[0][0]']
ormalization)
batch_normalization_101 (Batch (None, 25, 25, 64) 192 ['conv2d_101[0][0]']
Normalization)
batch_normalization_104 (Batch (None, 25, 25, 96) 288 ['conv2d_104[0][0]']
Normalization)
batch_normalization_105 (Batch (None, 25, 25, 32) 96 ['conv2d_105[0][0]']
Normalization)
activation_99 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_99[0][0]']
activation_101 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_101[0][0]']
activation_104 (Activation) (None, 25, 25, 96) 0 ['batch_normalization_104[0][0]']
activation_105 (Activation) (None, 25, 25, 32) 0 ['batch_normalization_105[0][0]']
mixed0 (Concatenate) (None, 25, 25, 256) 0 ['activation_99[0][0]',
'activation_101[0][0]',
'activation_104[0][0]',
'activation_105[0][0]']
conv2d_109 (Conv2D) (None, 25, 25, 64) 16384 ['mixed0[0][0]']
batch_normalization_109 (Batch (None, 25, 25, 64) 192 ['conv2d_109[0][0]']
Normalization)
activation_109 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_109[0][0]']
conv2d_107 (Conv2D) (None, 25, 25, 48) 12288 ['mixed0[0][0]']
conv2d_110 (Conv2D) (None, 25, 25, 96) 55296 ['activation_109[0][0]']
batch_normalization_107 (Batch (None, 25, 25, 48) 144 ['conv2d_107[0][0]']
Normalization)
batch_normalization_110 (Batch (None, 25, 25, 96) 288 ['conv2d_110[0][0]']
Normalization)
activation_107 (Activation) (None, 25, 25, 48) 0 ['batch_normalization_107[0][0]']
activation_110 (Activation) (None, 25, 25, 96) 0 ['batch_normalization_110[0][0]']
average_pooling2d_10 (AverageP (None, 25, 25, 256) 0 ['mixed0[0][0]']
ooling2D)
conv2d_106 (Conv2D) (None, 25, 25, 64) 16384 ['mixed0[0][0]']
conv2d_108 (Conv2D) (None, 25, 25, 64) 76800 ['activation_107[0][0]']
conv2d_111 (Conv2D) (None, 25, 25, 96) 82944 ['activation_110[0][0]']
conv2d_112 (Conv2D) (None, 25, 25, 64) 16384 ['average_pooling2d_10[0][0]']
batch_normalization_106 (Batch (None, 25, 25, 64) 192 ['conv2d_106[0][0]']
Normalization)
batch_normalization_108 (Batch (None, 25, 25, 64) 192 ['conv2d_108[0][0]']
Normalization)
batch_normalization_111 (Batch (None, 25, 25, 96) 288 ['conv2d_111[0][0]']
Normalization)
batch_normalization_112 (Batch (None, 25, 25, 64) 192 ['conv2d_112[0][0]']
Normalization)
activation_106 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_106[0][0]']
activation_108 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_108[0][0]']
activation_111 (Activation) (None, 25, 25, 96) 0 ['batch_normalization_111[0][0]']
activation_112 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_112[0][0]']
mixed1 (Concatenate) (None, 25, 25, 288) 0 ['activation_106[0][0]',
'activation_108[0][0]',
'activation_111[0][0]',
'activation_112[0][0]']
conv2d_116 (Conv2D) (None, 25, 25, 64) 18432 ['mixed1[0][0]']
batch_normalization_116 (Batch (None, 25, 25, 64) 192 ['conv2d_116[0][0]']
Normalization)
activation_116 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_116[0][0]']
conv2d_114 (Conv2D) (None, 25, 25, 48) 13824 ['mixed1[0][0]']
conv2d_117 (Conv2D) (None, 25, 25, 96) 55296 ['activation_116[0][0]']
batch_normalization_114 (Batch (None, 25, 25, 48) 144 ['conv2d_114[0][0]']
Normalization)
batch_normalization_117 (Batch (None, 25, 25, 96) 288 ['conv2d_117[0][0]']
Normalization)
activation_114 (Activation) (None, 25, 25, 48) 0 ['batch_normalization_114[0][0]']
activation_117 (Activation) (None, 25, 25, 96) 0 ['batch_normalization_117[0][0]']
average_pooling2d_11 (AverageP (None, 25, 25, 288) 0 ['mixed1[0][0]']
ooling2D)
conv2d_113 (Conv2D) (None, 25, 25, 64) 18432 ['mixed1[0][0]']
conv2d_115 (Conv2D) (None, 25, 25, 64) 76800 ['activation_114[0][0]']
conv2d_118 (Conv2D) (None, 25, 25, 96) 82944 ['activation_117[0][0]']
conv2d_119 (Conv2D) (None, 25, 25, 64) 18432 ['average_pooling2d_11[0][0]']
batch_normalization_113 (Batch (None, 25, 25, 64) 192 ['conv2d_113[0][0]']
Normalization)
batch_normalization_115 (Batch (None, 25, 25, 64) 192 ['conv2d_115[0][0]']
Normalization)
batch_normalization_118 (Batch (None, 25, 25, 96) 288 ['conv2d_118[0][0]']
Normalization)
batch_normalization_119 (Batch (None, 25, 25, 64) 192 ['conv2d_119[0][0]']
Normalization)
activation_113 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_113[0][0]']
activation_115 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_115[0][0]']
activation_118 (Activation) (None, 25, 25, 96) 0 ['batch_normalization_118[0][0]']
activation_119 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_119[0][0]']
mixed2 (Concatenate) (None, 25, 25, 288) 0 ['activation_113[0][0]',
'activation_115[0][0]',
'activation_118[0][0]',
'activation_119[0][0]']
conv2d_121 (Conv2D) (None, 25, 25, 64) 18432 ['mixed2[0][0]']
batch_normalization_121 (Batch (None, 25, 25, 64) 192 ['conv2d_121[0][0]']
Normalization)
activation_121 (Activation) (None, 25, 25, 64) 0 ['batch_normalization_121[0][0]']
conv2d_122 (Conv2D) (None, 25, 25, 96) 55296 ['activation_121[0][0]']
batch_normalization_122 (Batch (None, 25, 25, 96) 288 ['conv2d_122[0][0]']
Normalization)
activation_122 (Activation) (None, 25, 25, 96) 0 ['batch_normalization_122[0][0]']
conv2d_120 (Conv2D) (None, 12, 12, 384) 995328 ['mixed2[0][0]']
conv2d_123 (Conv2D) (None, 12, 12, 96) 82944 ['activation_122[0][0]']
batch_normalization_120 (Batch (None, 12, 12, 384) 1152 ['conv2d_120[0][0]']
Normalization)
batch_normalization_123 (Batch (None, 12, 12, 96) 288 ['conv2d_123[0][0]']
Normalization)
activation_120 (Activation) (None, 12, 12, 384) 0 ['batch_normalization_120[0][0]']
activation_123 (Activation) (None, 12, 12, 96) 0 ['batch_normalization_123[0][0]']
max_pooling2d_6 (MaxPooling2D) (None, 12, 12, 288) 0 ['mixed2[0][0]']
mixed3 (Concatenate) (None, 12, 12, 768) 0 ['activation_120[0][0]',
'activation_123[0][0]',
'max_pooling2d_6[0][0]']
conv2d_128 (Conv2D) (None, 12, 12, 128) 98304 ['mixed3[0][0]']
batch_normalization_128 (Batch (None, 12, 12, 128) 384 ['conv2d_128[0][0]']
Normalization)
activation_128 (Activation) (None, 12, 12, 128) 0 ['batch_normalization_128[0][0]']
conv2d_129 (Conv2D) (None, 12, 12, 128) 114688 ['activation_128[0][0]']
batch_normalization_129 (Batch (None, 12, 12, 128) 384 ['conv2d_129[0][0]']
Normalization)
activation_129 (Activation) (None, 12, 12, 128) 0 ['batch_normalization_129[0][0]']
conv2d_125 (Conv2D) (None, 12, 12, 128) 98304 ['mixed3[0][0]']
conv2d_130 (Conv2D) (None, 12, 12, 128) 114688 ['activation_129[0][0]']
batch_normalization_125 (Batch (None, 12, 12, 128) 384 ['conv2d_125[0][0]']
Normalization)
batch_normalization_130 (Batch (None, 12, 12, 128) 384 ['conv2d_130[0][0]']
Normalization)
activation_125 (Activation) (None, 12, 12, 128) 0 ['batch_normalization_125[0][0]']
activation_130 (Activation) (None, 12, 12, 128) 0 ['batch_normalization_130[0][0]']
conv2d_126 (Conv2D) (None, 12, 12, 128) 114688 ['activation_125[0][0]']
conv2d_131 (Conv2D) (None, 12, 12, 128) 114688 ['activation_130[0][0]']
batch_normalization_126 (Batch (None, 12, 12, 128) 384 ['conv2d_126[0][0]']
Normalization)
batch_normalization_131 (Batch (None, 12, 12, 128) 384 ['conv2d_131[0][0]']
Normalization)
activation_126 (Activation) (None, 12, 12, 128) 0 ['batch_normalization_126[0][0]']
activation_131 (Activation) (None, 12, 12, 128) 0 ['batch_normalization_131[0][0]']
average_pooling2d_12 (AverageP (None, 12, 12, 768) 0 ['mixed3[0][0]']
ooling2D)
conv2d_124 (Conv2D) (None, 12, 12, 192) 147456 ['mixed3[0][0]']
conv2d_127 (Conv2D) (None, 12, 12, 192) 172032 ['activation_126[0][0]']
conv2d_132 (Conv2D) (None, 12, 12, 192) 172032 ['activation_131[0][0]']
conv2d_133 (Conv2D) (None, 12, 12, 192) 147456 ['average_pooling2d_12[0][0]']
batch_normalization_124 (Batch (None, 12, 12, 192) 576 ['conv2d_124[0][0]']
Normalization)
batch_normalization_127 (Batch (None, 12, 12, 192) 576 ['conv2d_127[0][0]']
Normalization)
batch_normalization_132 (Batch (None, 12, 12, 192) 576 ['conv2d_132[0][0]']
Normalization)
batch_normalization_133 (Batch (None, 12, 12, 192) 576 ['conv2d_133[0][0]']
Normalization)
activation_124 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_124[0][0]']
activation_127 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_127[0][0]']
activation_132 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_132[0][0]']
activation_133 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_133[0][0]']
mixed4 (Concatenate) (None, 12, 12, 768) 0 ['activation_124[0][0]',
'activation_127[0][0]',
'activation_132[0][0]',
'activation_133[0][0]']
conv2d_138 (Conv2D) (None, 12, 12, 160) 122880 ['mixed4[0][0]']
batch_normalization_138 (Batch (None, 12, 12, 160) 480 ['conv2d_138[0][0]']
Normalization)
activation_138 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_138[0][0]']
conv2d_139 (Conv2D) (None, 12, 12, 160) 179200 ['activation_138[0][0]']
batch_normalization_139 (Batch (None, 12, 12, 160) 480 ['conv2d_139[0][0]']
Normalization)
activation_139 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_139[0][0]']
conv2d_135 (Conv2D) (None, 12, 12, 160) 122880 ['mixed4[0][0]']
conv2d_140 (Conv2D) (None, 12, 12, 160) 179200 ['activation_139[0][0]']
batch_normalization_135 (Batch (None, 12, 12, 160) 480 ['conv2d_135[0][0]']
Normalization)
batch_normalization_140 (Batch (None, 12, 12, 160) 480 ['conv2d_140[0][0]']
Normalization)
activation_135 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_135[0][0]']
activation_140 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_140[0][0]']
conv2d_136 (Conv2D) (None, 12, 12, 160) 179200 ['activation_135[0][0]']
conv2d_141 (Conv2D) (None, 12, 12, 160) 179200 ['activation_140[0][0]']
batch_normalization_136 (Batch (None, 12, 12, 160) 480 ['conv2d_136[0][0]']
Normalization)
batch_normalization_141 (Batch (None, 12, 12, 160) 480 ['conv2d_141[0][0]']
Normalization)
activation_136 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_136[0][0]']
activation_141 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_141[0][0]']
average_pooling2d_13 (AverageP (None, 12, 12, 768) 0 ['mixed4[0][0]']
ooling2D)
conv2d_134 (Conv2D) (None, 12, 12, 192) 147456 ['mixed4[0][0]']
conv2d_137 (Conv2D) (None, 12, 12, 192) 215040 ['activation_136[0][0]']
conv2d_142 (Conv2D) (None, 12, 12, 192) 215040 ['activation_141[0][0]']
conv2d_143 (Conv2D) (None, 12, 12, 192) 147456 ['average_pooling2d_13[0][0]']
batch_normalization_134 (Batch (None, 12, 12, 192) 576 ['conv2d_134[0][0]']
Normalization)
batch_normalization_137 (Batch (None, 12, 12, 192) 576 ['conv2d_137[0][0]']
Normalization)
batch_normalization_142 (Batch (None, 12, 12, 192) 576 ['conv2d_142[0][0]']
Normalization)
batch_normalization_143 (Batch (None, 12, 12, 192) 576 ['conv2d_143[0][0]']
Normalization)
activation_134 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_134[0][0]']
activation_137 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_137[0][0]']
activation_142 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_142[0][0]']
activation_143 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_143[0][0]']
mixed5 (Concatenate) (None, 12, 12, 768) 0 ['activation_134[0][0]',
'activation_137[0][0]',
'activation_142[0][0]',
'activation_143[0][0]']
conv2d_148 (Conv2D) (None, 12, 12, 160) 122880 ['mixed5[0][0]']
batch_normalization_148 (Batch (None, 12, 12, 160) 480 ['conv2d_148[0][0]']
Normalization)
activation_148 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_148[0][0]']
conv2d_149 (Conv2D) (None, 12, 12, 160) 179200 ['activation_148[0][0]']
batch_normalization_149 (Batch (None, 12, 12, 160) 480 ['conv2d_149[0][0]']
Normalization)
activation_149 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_149[0][0]']
conv2d_145 (Conv2D) (None, 12, 12, 160) 122880 ['mixed5[0][0]']
conv2d_150 (Conv2D) (None, 12, 12, 160) 179200 ['activation_149[0][0]']
batch_normalization_145 (Batch (None, 12, 12, 160) 480 ['conv2d_145[0][0]']
Normalization)
batch_normalization_150 (Batch (None, 12, 12, 160) 480 ['conv2d_150[0][0]']
Normalization)
activation_145 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_145[0][0]']
activation_150 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_150[0][0]']
conv2d_146 (Conv2D) (None, 12, 12, 160) 179200 ['activation_145[0][0]']
conv2d_151 (Conv2D) (None, 12, 12, 160) 179200 ['activation_150[0][0]']
batch_normalization_146 (Batch (None, 12, 12, 160) 480 ['conv2d_146[0][0]']
Normalization)
batch_normalization_151 (Batch (None, 12, 12, 160) 480 ['conv2d_151[0][0]']
Normalization)
activation_146 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_146[0][0]']
activation_151 (Activation) (None, 12, 12, 160) 0 ['batch_normalization_151[0][0]']
average_pooling2d_14 (AverageP (None, 12, 12, 768) 0 ['mixed5[0][0]']
ooling2D)
conv2d_144 (Conv2D) (None, 12, 12, 192) 147456 ['mixed5[0][0]']
conv2d_147 (Conv2D) (None, 12, 12, 192) 215040 ['activation_146[0][0]']
conv2d_152 (Conv2D) (None, 12, 12, 192) 215040 ['activation_151[0][0]']
conv2d_153 (Conv2D) (None, 12, 12, 192) 147456 ['average_pooling2d_14[0][0]']
batch_normalization_144 (Batch (None, 12, 12, 192) 576 ['conv2d_144[0][0]']
Normalization)
batch_normalization_147 (Batch (None, 12, 12, 192) 576 ['conv2d_147[0][0]']
Normalization)
batch_normalization_152 (Batch (None, 12, 12, 192) 576 ['conv2d_152[0][0]']
Normalization)
batch_normalization_153 (Batch (None, 12, 12, 192) 576 ['conv2d_153[0][0]']
Normalization)
activation_144 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_144[0][0]']
activation_147 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_147[0][0]']
activation_152 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_152[0][0]']
activation_153 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_153[0][0]']
mixed6 (Concatenate) (None, 12, 12, 768) 0 ['activation_144[0][0]',
'activation_147[0][0]',
'activation_152[0][0]',
'activation_153[0][0]']
conv2d_158 (Conv2D) (None, 12, 12, 192) 147456 ['mixed6[0][0]']
batch_normalization_158 (Batch (None, 12, 12, 192) 576 ['conv2d_158[0][0]']
Normalization)
activation_158 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_158[0][0]']
conv2d_159 (Conv2D) (None, 12, 12, 192) 258048 ['activation_158[0][0]']
batch_normalization_159 (Batch (None, 12, 12, 192) 576 ['conv2d_159[0][0]']
Normalization)
activation_159 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_159[0][0]']
conv2d_155 (Conv2D) (None, 12, 12, 192) 147456 ['mixed6[0][0]']
conv2d_160 (Conv2D) (None, 12, 12, 192) 258048 ['activation_159[0][0]']
batch_normalization_155 (Batch (None, 12, 12, 192) 576 ['conv2d_155[0][0]']
Normalization)
batch_normalization_160 (Batch (None, 12, 12, 192) 576 ['conv2d_160[0][0]']
Normalization)
activation_155 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_155[0][0]']
activation_160 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_160[0][0]']
conv2d_156 (Conv2D) (None, 12, 12, 192) 258048 ['activation_155[0][0]']
conv2d_161 (Conv2D) (None, 12, 12, 192) 258048 ['activation_160[0][0]']
batch_normalization_156 (Batch (None, 12, 12, 192) 576 ['conv2d_156[0][0]']
Normalization)
batch_normalization_161 (Batch (None, 12, 12, 192) 576 ['conv2d_161[0][0]']
Normalization)
activation_156 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_156[0][0]']
activation_161 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_161[0][0]']
average_pooling2d_15 (AverageP (None, 12, 12, 768) 0 ['mixed6[0][0]']
ooling2D)
conv2d_154 (Conv2D) (None, 12, 12, 192) 147456 ['mixed6[0][0]']
conv2d_157 (Conv2D) (None, 12, 12, 192) 258048 ['activation_156[0][0]']
conv2d_162 (Conv2D) (None, 12, 12, 192) 258048 ['activation_161[0][0]']
conv2d_163 (Conv2D) (None, 12, 12, 192) 147456 ['average_pooling2d_15[0][0]']
batch_normalization_154 (Batch (None, 12, 12, 192) 576 ['conv2d_154[0][0]']
Normalization)
batch_normalization_157 (Batch (None, 12, 12, 192) 576 ['conv2d_157[0][0]']
Normalization)
batch_normalization_162 (Batch (None, 12, 12, 192) 576 ['conv2d_162[0][0]']
Normalization)
batch_normalization_163 (Batch (None, 12, 12, 192) 576 ['conv2d_163[0][0]']
Normalization)
activation_154 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_154[0][0]']
activation_157 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_157[0][0]']
activation_162 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_162[0][0]']
activation_163 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_163[0][0]']
mixed7 (Concatenate) (None, 12, 12, 768) 0 ['activation_154[0][0]',
'activation_157[0][0]',
'activation_162[0][0]',
'activation_163[0][0]']
conv2d_166 (Conv2D) (None, 12, 12, 192) 147456 ['mixed7[0][0]']
batch_normalization_166 (Batch (None, 12, 12, 192) 576 ['conv2d_166[0][0]']
Normalization)
activation_166 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_166[0][0]']
conv2d_167 (Conv2D) (None, 12, 12, 192) 258048 ['activation_166[0][0]']
batch_normalization_167 (Batch (None, 12, 12, 192) 576 ['conv2d_167[0][0]']
Normalization)
activation_167 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_167[0][0]']
conv2d_164 (Conv2D) (None, 12, 12, 192) 147456 ['mixed7[0][0]']
conv2d_168 (Conv2D) (None, 12, 12, 192) 258048 ['activation_167[0][0]']
batch_normalization_164 (Batch (None, 12, 12, 192) 576 ['conv2d_164[0][0]']
Normalization)
batch_normalization_168 (Batch (None, 12, 12, 192) 576 ['conv2d_168[0][0]']
Normalization)
activation_164 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_164[0][0]']
activation_168 (Activation) (None, 12, 12, 192) 0 ['batch_normalization_168[0][0]']
conv2d_165 (Conv2D) (None, 5, 5, 320) 552960 ['activation_164[0][0]']
conv2d_169 (Conv2D) (None, 5, 5, 192) 331776 ['activation_168[0][0]']
batch_normalization_165 (Batch (None, 5, 5, 320) 960 ['conv2d_165[0][0]']
Normalization)
batch_normalization_169 (Batch (None, 5, 5, 192) 576 ['conv2d_169[0][0]']
Normalization)
activation_165 (Activation) (None, 5, 5, 320) 0 ['batch_normalization_165[0][0]']
activation_169 (Activation) (None, 5, 5, 192) 0 ['batch_normalization_169[0][0]']
max_pooling2d_7 (MaxPooling2D) (None, 5, 5, 768) 0 ['mixed7[0][0]']
mixed8 (Concatenate) (None, 5, 5, 1280) 0 ['activation_165[0][0]',
'activation_169[0][0]',
'max_pooling2d_7[0][0]']
conv2d_174 (Conv2D) (None, 5, 5, 448) 573440 ['mixed8[0][0]']
batch_normalization_174 (Batch (None, 5, 5, 448) 1344 ['conv2d_174[0][0]']
Normalization)
activation_174 (Activation) (None, 5, 5, 448) 0 ['batch_normalization_174[0][0]']
conv2d_171 (Conv2D) (None, 5, 5, 384) 491520 ['mixed8[0][0]']
conv2d_175 (Conv2D) (None, 5, 5, 384) 1548288 ['activation_174[0][0]']
batch_normalization_171 (Batch (None, 5, 5, 384) 1152 ['conv2d_171[0][0]']
Normalization)
batch_normalization_175 (Batch (None, 5, 5, 384) 1152 ['conv2d_175[0][0]']
Normalization)
activation_171 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_171[0][0]']
activation_175 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_175[0][0]']
conv2d_172 (Conv2D) (None, 5, 5, 384) 442368 ['activation_171[0][0]']
conv2d_173 (Conv2D) (None, 5, 5, 384) 442368 ['activation_171[0][0]']
conv2d_176 (Conv2D) (None, 5, 5, 384) 442368 ['activation_175[0][0]']
conv2d_177 (Conv2D) (None, 5, 5, 384) 442368 ['activation_175[0][0]']
average_pooling2d_16 (AverageP (None, 5, 5, 1280) 0 ['mixed8[0][0]']
ooling2D)
conv2d_170 (Conv2D) (None, 5, 5, 320) 409600 ['mixed8[0][0]']
batch_normalization_172 (Batch (None, 5, 5, 384) 1152 ['conv2d_172[0][0]']
Normalization)
batch_normalization_173 (Batch (None, 5, 5, 384) 1152 ['conv2d_173[0][0]']
Normalization)
batch_normalization_176 (Batch (None, 5, 5, 384) 1152 ['conv2d_176[0][0]']
Normalization)
batch_normalization_177 (Batch (None, 5, 5, 384) 1152 ['conv2d_177[0][0]']
Normalization)
conv2d_178 (Conv2D) (None, 5, 5, 192) 245760 ['average_pooling2d_16[0][0]']
batch_normalization_170 (Batch (None, 5, 5, 320) 960 ['conv2d_170[0][0]']
Normalization)
activation_172 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_172[0][0]']
activation_173 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_173[0][0]']
activation_176 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_176[0][0]']
activation_177 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_177[0][0]']
batch_normalization_178 (Batch (None, 5, 5, 192) 576 ['conv2d_178[0][0]']
Normalization)
activation_170 (Activation) (None, 5, 5, 320) 0 ['batch_normalization_170[0][0]']
mixed9_0 (Concatenate) (None, 5, 5, 768) 0 ['activation_172[0][0]',
'activation_173[0][0]']
concatenate_2 (Concatenate) (None, 5, 5, 768) 0 ['activation_176[0][0]',
'activation_177[0][0]']
activation_178 (Activation) (None, 5, 5, 192) 0 ['batch_normalization_178[0][0]']
mixed9 (Concatenate) (None, 5, 5, 2048) 0 ['activation_170[0][0]',
'mixed9_0[0][0]',
'concatenate_2[0][0]',
'activation_178[0][0]']
conv2d_183 (Conv2D) (None, 5, 5, 448) 917504 ['mixed9[0][0]']
batch_normalization_183 (Batch (None, 5, 5, 448) 1344 ['conv2d_183[0][0]']
Normalization)
activation_183 (Activation) (None, 5, 5, 448) 0 ['batch_normalization_183[0][0]']
conv2d_180 (Conv2D) (None, 5, 5, 384) 786432 ['mixed9[0][0]']
conv2d_184 (Conv2D) (None, 5, 5, 384) 1548288 ['activation_183[0][0]']
batch_normalization_180 (Batch (None, 5, 5, 384) 1152 ['conv2d_180[0][0]']
Normalization)
batch_normalization_184 (Batch (None, 5, 5, 384) 1152 ['conv2d_184[0][0]']
Normalization)
activation_180 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_180[0][0]']
activation_184 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_184[0][0]']
conv2d_181 (Conv2D) (None, 5, 5, 384) 442368 ['activation_180[0][0]']
conv2d_182 (Conv2D) (None, 5, 5, 384) 442368 ['activation_180[0][0]']
conv2d_185 (Conv2D) (None, 5, 5, 384) 442368 ['activation_184[0][0]']
conv2d_186 (Conv2D) (None, 5, 5, 384) 442368 ['activation_184[0][0]']
average_pooling2d_17 (AverageP (None, 5, 5, 2048) 0 ['mixed9[0][0]']
ooling2D)
conv2d_179 (Conv2D) (None, 5, 5, 320) 655360 ['mixed9[0][0]']
batch_normalization_181 (Batch (None, 5, 5, 384) 1152 ['conv2d_181[0][0]']
Normalization)
batch_normalization_182 (Batch (None, 5, 5, 384) 1152 ['conv2d_182[0][0]']
Normalization)
batch_normalization_185 (Batch (None, 5, 5, 384) 1152 ['conv2d_185[0][0]']
Normalization)
batch_normalization_186 (Batch (None, 5, 5, 384) 1152 ['conv2d_186[0][0]']
Normalization)
conv2d_187 (Conv2D) (None, 5, 5, 192) 393216 ['average_pooling2d_17[0][0]']
batch_normalization_179 (Batch (None, 5, 5, 320) 960 ['conv2d_179[0][0]']
Normalization)
activation_181 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_181[0][0]']
activation_182 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_182[0][0]']
activation_185 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_185[0][0]']
activation_186 (Activation) (None, 5, 5, 384) 0 ['batch_normalization_186[0][0]']
batch_normalization_187 (Batch (None, 5, 5, 192) 576 ['conv2d_187[0][0]']
Normalization)
activation_179 (Activation) (None, 5, 5, 320) 0 ['batch_normalization_179[0][0]']
mixed9_1 (Concatenate) (None, 5, 5, 768) 0 ['activation_181[0][0]',
'activation_182[0][0]']
concatenate_3 (Concatenate) (None, 5, 5, 768) 0 ['activation_185[0][0]',
'activation_186[0][0]']
activation_187 (Activation) (None, 5, 5, 192) 0 ['batch_normalization_187[0][0]']
mixed10 (Concatenate) (None, 5, 5, 2048) 0 ['activation_179[0][0]',
'mixed9_1[0][0]',
'concatenate_3[0][0]',
'activation_187[0][0]']
flatten_1 (Flatten) (None, 51200) 0 ['mixed10[0][0]']
dense_1 (Dense) (None, 10) 512010 ['flatten_1[0][0]']
==================================================================================================
Total params: 22,314,794
Trainable params: 512,010
Non-trainable params: 21,802,784
__________________________________________________________________________________________________
Train model¶
In [ ]:
# Defining the cost and model optimization method to use
model.compile(
loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy']
)
In [ ]:
# Using the Image Data Generator to import the images from the dataset
from tensorflow.keras.preprocessing.image import ImageDataGenerator
train_datagen = ImageDataGenerator(rescale = 1./255,
shear_range = 0.2,
zoom_range = 0.2,
horizontal_flip = True)
test_datagen = ImageDataGenerator(rescale = 1./255)
In [ ]:
# Training Generator
training_set = train_datagen.flow_from_directory('/content/gdrive/My Drive/Colab Notebooks/dataset8020/train',
target_size = (224, 224),
batch_size = 32,
class_mode = 'categorical')
Found 8000 images belonging to 10 classes.
In [ ]:
# Testing Generator
test_set = test_datagen.flow_from_directory('/content/gdrive/My Drive/Colab Notebooks/dataset8020/val',
target_size = (224, 224),
batch_size = 32,
class_mode = 'categorical')
Found 2000 images belonging to 10 classes.
In [ ]:
# fit the model, it will take some time to execute
r = model.fit(
training_set,
validation_data=test_set,
epochs=5,
steps_per_epoch=len(training_set),
validation_steps=len(test_set)
)
Epoch 1/5 250/250 [==============================] - 1502s 6s/step - loss: 4.0302 - accuracy: 0.6111 - val_loss: 2.2804 - val_accuracy: 0.7075 Epoch 2/5 250/250 [==============================] - 1007s 4s/step - loss: 1.8126 - accuracy: 0.7675 - val_loss: 1.3260 - val_accuracy: 0.8045 Epoch 3/5 250/250 [==============================] - 1015s 4s/step - loss: 1.8028 - accuracy: 0.7970 - val_loss: 2.2641 - val_accuracy: 0.7620 Epoch 4/5 250/250 [==============================] - 1022s 4s/step - loss: 1.5416 - accuracy: 0.8234 - val_loss: 0.9595 - val_accuracy: 0.8770 Epoch 5/5 250/250 [==============================] - 1020s 4s/step - loss: 1.5698 - accuracy: 0.8371 - val_loss: 1.3281 - val_accuracy: 0.8495
Visualization Loss And Accuracy¶
In [ ]:
# Plot the Loss and Accuracy
# Loss
model.evaluate(test_set)
plt.plot(r.history['loss'], label='train loss')
plt.plot(r.history['val_loss'], label='val loss')
plt.legend()
plt.show()
# Accuracy
plt.plot(r.history['accuracy'], label='train acc')
plt.plot(r.history['val_accuracy'], label='val acc')
plt.legend()
plt.show()
63/63 [==============================] - 188s 3s/step - loss: 1.3281 - accuracy: 0.8495
Menyimpan Model¶
In [ ]:
# Saving the model as a h5 file
from tensorflow.keras.models import load_model
model.save('model_inception.h5')
Load Model¶
In [ ]:
model = load_model('model_inception.h5')
--------------------------------------------------------------------------- OSError Traceback (most recent call last) <ipython-input-14-6a079f9de6f3> in <module> ----> 1 model = load_model('model_inception.h5') /usr/local/lib/python3.7/dist-packages/keras/utils/traceback_utils.py in error_handler(*args, **kwargs) 65 except Exception as e: # pylint: disable=broad-except 66 filtered_tb = _process_traceback_frames(e.__traceback__) ---> 67 raise e.with_traceback(filtered_tb) from None 68 finally: 69 del filtered_tb /usr/local/lib/python3.7/dist-packages/keras/saving/save.py in load_model(filepath, custom_objects, compile, options) 204 if isinstance(filepath_str, str): 205 if not tf.io.gfile.exists(filepath_str): --> 206 raise IOError(f'No file or directory found at {filepath_str}') 207 208 if tf.io.gfile.isdir(filepath_str): OSError: No file or directory found at model_inception.h5
Predict¶
In [ ]:
y_pred = model.predict(test_set)
63/63 [==============================] - 195s 3s/step
In [ ]:
y_pred
Out[ ]:
array([[0.0000000e+00, 1.9883246e-34, 8.9775664e-01, ..., 1.0224147e-01,
2.9612454e-15, 1.8457501e-06],
[1.0475146e-31, 2.4344494e-23, 1.2251389e-19, ..., 9.9999994e-01,
1.2092257e-21, 6.1555072e-10],
[6.6823721e-15, 1.9258282e-06, 2.0794512e-06, ..., 3.5903615e-01,
6.8596691e-11, 2.8802851e-01],
...,
[9.9999994e-01, 1.1173088e-09, 7.8183576e-19, ..., 9.5940103e-11,
4.4835802e-14, 5.0567840e-16],
[9.4067271e-14, 3.9219127e-18, 3.4441447e-07, ..., 9.9999624e-01,
3.0819001e-06, 4.1683735e-20],
[1.0242016e-26, 5.1891837e-12, 5.6711569e-10, ..., 1.2128113e-14,
3.9022973e-15, 1.2875653e-19]], dtype=float32)
In [ ]:
y_pred = np.argmax(y_pred, axis=1)
In [ ]:
y_pred
Out[ ]:
array([2, 7, 7, ..., 0, 7, 6])
In [ ]:
loss_final, accuracy_final = model.evaluate(test_set)
63/63 [==============================] - 189s 3s/step - loss: 1.3281 - accuracy: 0.8495
In [ ]:
print("Final loss: {:.2f}".format(loss_final))
print("Final accuracy: {:.2f}".format(accuracy_final))
Final loss: 1.33 Final accuracy: 0.85
Predict Insert Image¶
In [ ]:
%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np
from tensorflow.keras.preprocessing import image
from google.colab import files
uploaded = files.upload()
for fn in uploaded.keys():
# predicting images
path = '/content/' + fn
img = image.load_img(path, target_size=(224, 224))
x = image.img_to_array(img)
plt.imshow(x/255.)
x = np.expand_dims(x, axis=0)
images = np.vstack([x])
classes = model.predict(images, batch_size=32)
index = tf.argmax(classes[0], axis=0)
print(classes[0])
if classes[0][0]==1:
print("Bacterial spot")
elif classes[0][1]==1:
print("Early blight")
elif classes[0][2]==1:
print("Late blight")
elif classes[0][3]==1:
print("Leaf Mold")
elif classes[0][4]==1:
print("Septoria Leaf Spot")
elif classes[0][5]==1:
print("Spider Mites")
elif classes[0][6]==1:
print("Target Spot")
elif classes[0][7]==1:
print("Tomato Yellow Leaf Curl Virus")
elif classes[0][8]==1:
print("Tomato Mosaic Virus")
else:
print("Healthy")
Saving mosaic.jpg to mosaic (1).jpg 1/1 [==============================] - 0s 144ms/step 3 Leaf Mold
Klasifikasi PEDANTO VGG16¶
Modules dan Packages¶
In [ ]:
# Basic Imports
import numpy as np
import matplotlib.pyplot as plt
from glob import glob
from PIL import Image
# Imports untuk membangun CNN
from keras.layers import Input, Lambda, Dense, Flatten
from keras.models import Model
from keras.applications.vgg16 import VGG16
from keras.applications.vgg16 import preprocess_input
from keras.preprocessing import image
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential, load_model
# Ignore Warnings
import warnings
warnings.filterwarnings("ignore")
Import Dataset¶
In [ ]:
# Membuat semua foto ter-resize pada ukuran 224x224
IMAGE_SIZE = [224, 224]
categories = ["Tomato___Bacterial_spot", "Tomato___Early_blight", "Tomato___Healthy", "Tomato___Late_blight",
"Tomato___Leaf_Mold", "Tomato___Mosaic_virus", "Tomato___Septoria_leaf_spot", "Tomato___Spider_mites", "Tomato___Target_Spot", "Tomato___Yellow_Leaf_Curl_Virus"]
# Mencari file yang terkait pada folder dataset
train_path = '/content/gdrive/MyDrive/Colab Notebooks/dataset8020/train/'
test_path = '/content/gdrive/MyDrive/Colab Notebooks/dataset8020/val/'
Data Preprocessing¶
In [ ]:
def normalize(input_image, input_mask):
input_image = tf.cast(input_image, tf.float32) / 255.0
input_mask -= 1
return input_image, input_mask
VGG16¶
VGG16 Keras Model untuk Klasifikasi Gambar¶
In [ ]:
# Menambahkan Layer untuk Preprocessing pada VGG
vgg = VGG16(input_shape = IMAGE_SIZE + [3],
weights = 'imagenet',
include_top=False)
In [ ]:
# Untuk perintah menolak men training data yang telah digunakan sebelumnya
for layer in vgg.layers:
layer.trainable = False
In [ ]:
# Folders in the Training Set
folders = glob('/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/*')
folders
Out[ ]:
['/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Spider_mites', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Septoria_leaf_spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Late_blight', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Mosaic_virus', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Leaf_Mold', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Healthy', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Bacterial_spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Early_blight', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Target_Spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Yellow_Leaf_Curl_Virus']
Menambah Layer pada Model VGG16¶
In [ ]:
x = Flatten()(vgg.output)
x = Dense(1000, activation='relu')(x)
# Lapisan Terakhir dengan aktivasi softmax untuk output biner
prediction = Dense(len(categories), activation='softmax')(x)
In [ ]:
# Membuat obbjek model
model = Model(inputs=vgg.input, outputs=prediction)
# Melihat struktur model
model.summary()
Model: "model_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_2 (InputLayer) [(None, 224, 224, 3)] 0
block1_conv1 (Conv2D) (None, 224, 224, 64) 1792
block1_conv2 (Conv2D) (None, 224, 224, 64) 36928
block1_pool (MaxPooling2D) (None, 112, 112, 64) 0
block2_conv1 (Conv2D) (None, 112, 112, 128) 73856
block2_conv2 (Conv2D) (None, 112, 112, 128) 147584
block2_pool (MaxPooling2D) (None, 56, 56, 128) 0
block3_conv1 (Conv2D) (None, 56, 56, 256) 295168
block3_conv2 (Conv2D) (None, 56, 56, 256) 590080
block3_conv3 (Conv2D) (None, 56, 56, 256) 590080
block3_pool (MaxPooling2D) (None, 28, 28, 256) 0
block4_conv1 (Conv2D) (None, 28, 28, 512) 1180160
block4_conv2 (Conv2D) (None, 28, 28, 512) 2359808
block4_conv3 (Conv2D) (None, 28, 28, 512) 2359808
block4_pool (MaxPooling2D) (None, 14, 14, 512) 0
block5_conv1 (Conv2D) (None, 14, 14, 512) 2359808
block5_conv2 (Conv2D) (None, 14, 14, 512) 2359808
block5_conv3 (Conv2D) (None, 14, 14, 512) 2359808
block5_pool (MaxPooling2D) (None, 7, 7, 512) 0
flatten_1 (Flatten) (None, 25088) 0
dense_2 (Dense) (None, 1000) 25089000
dense_3 (Dense) (None, 10) 10010
=================================================================
Total params: 39,813,698
Trainable params: 25,099,010
Non-trainable params: 14,714,688
_________________________________________________________________
Compile Model¶
In [ ]:
# tell the model what cost and optimization method to use
model.compile(
loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy']
)
Preparing Image For Model¶
In [ ]:
train_datagen = ImageDataGenerator(rescale = 1./255,
shear_range = 0.2,
zoom_range = 0.2,
horizontal_flip = True)
test_datagen = ImageDataGenerator(rescale = 1./255)
In [ ]:
training_set = train_datagen.flow_from_directory(train_path,
target_size = (224, 224),
batch_size = 32,
class_mode = 'categorical')
test_set = test_datagen.flow_from_directory(test_path,
target_size = (224, 224),
batch_size = 32,
class_mode = 'categorical')
Found 8000 images belonging to 10 classes. Found 2000 images belonging to 10 classes.
Train Model¶
In [ ]:
# Mencocokan model
r = model.fit_generator(
training_set,
validation_data=test_set,
epochs=5,
steps_per_epoch=len(training_set),
validation_steps=len(test_set)
)
Epoch 1/5 250/250 [==============================] - 3828s 15s/step - loss: 1.4204 - accuracy: 0.6785 - val_loss: 0.5440 - val_accuracy: 0.8035 Epoch 2/5 250/250 [==============================] - 117s 469ms/step - loss: 0.5472 - accuracy: 0.8126 - val_loss: 0.4228 - val_accuracy: 0.8455 Epoch 3/5 250/250 [==============================] - 118s 470ms/step - loss: 0.3846 - accuracy: 0.8696 - val_loss: 0.3903 - val_accuracy: 0.8685 Epoch 4/5 250/250 [==============================] - 118s 472ms/step - loss: 0.3630 - accuracy: 0.8710 - val_loss: 0.3219 - val_accuracy: 0.8870 Epoch 5/5 250/250 [==============================] - 117s 468ms/step - loss: 0.3021 - accuracy: 0.8931 - val_loss: 0.2770 - val_accuracy: 0.9050
Visualization Loss And Accuracy¶
In [ ]:
# Plot the Loss and Accuracy
# Loss
model.evaluate(test_set)
plt.plot(r.history['loss'], label='train loss')
plt.plot(r.history['val_loss'], label='val loss')
plt.legend()
plt.show()
plt.savefig('LossVal_loss')
# Accuracy
plt.plot(r.history['accuracy'], label='train acc')
plt.plot(r.history['val_accuracy'], label='val acc')
plt.legend()
plt.show()
plt.savefig('AccVal_acc')
63/63 [==============================] - 10s 151ms/step - loss: 0.2770 - accuracy: 0.9050
<Figure size 432x288 with 0 Axes>
Menyimpan Model¶
In [ ]:
# Saving the model as a h5 file
from tensorflow.keras.models import load_model
model.save('vgg16_model.h5')
Load Model¶
In [ ]:
model = load_model('vgg16_model.h5')
Predict¶
In [ ]:
y_pred = model.predict(test_set)
63/63 [==============================] - 11s 168ms/step
In [ ]:
y_pred
Out[ ]:
array([[1.5418239e-04, 7.6563996e-03, 3.5058329e-05, ..., 5.9890544e-01,
6.0090553e-02, 5.9241982e-04],
[1.9848589e-09, 3.0925603e-06, 6.3380703e-08, ..., 1.1981122e-05,
8.5274069e-06, 2.9561184e-08],
[2.8317691e-05, 2.0588029e-03, 1.0515270e-06, ..., 1.4159037e-02,
1.1849840e-02, 1.9398154e-04],
...,
[4.0504841e-05, 2.2264179e-04, 2.0513899e-07, ..., 4.9343280e-04,
4.1385425e-05, 9.9880099e-01],
[5.0203088e-03, 2.8763310e-05, 4.9834182e-07, ..., 8.6154260e-02,
9.0686584e-01, 2.8586022e-05],
[3.1297377e-06, 9.3861883e-05, 1.5117957e-05, ..., 9.9851545e-03,
8.8743404e-02, 9.8697354e-05]], dtype=float32)
In [ ]:
y_pred = np.argmax(y_pred, axis=1)
In [ ]:
y_pred
Out[ ]:
array([7, 5, 4, ..., 9, 8, 4])
In [ ]:
loss_final, accuracy_final = model.evaluate(test_set)
63/63 [==============================] - 14s 219ms/step - loss: 0.2770 - accuracy: 0.9050
In [ ]:
print("Final loss: {:.2f}".format(loss_final))
print("Final accuracy: {:.2f}".format(accuracy_final))
Final loss: 0.28 Final accuracy: 0.90
Predict Insert Image¶
In [ ]:
from google.colab import drive
drive.mount('/content/drive')
Mounted at /content/drive
In [ ]:
%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf
from tensorflow.keras.preprocessing import image
from google.colab import files
uploaded = files.upload()
for fn in uploaded.keys():
# predicting images
path = '/content/' + fn
img = image.load_img(path, target_size=(224, 224))
x = image.img_to_array(img)
plt.imshow(x/255.)
x = np.expand_dims(x, axis=0)
images = np.vstack([x])
classes = model.predict(images, batch_size=32)
index = tf.argmax(classes[0], axis=0)
print(classes[0])
if classes[0][0]==1:
print("Bacterial spot")
elif classes[0][1]==1:
print("Early blight")
elif classes[0][2]==1:
print("Late blight")
elif classes[0][3]==1:
print("Leaf Mold")
elif classes[0][4]==1:
print("Septoria Leaf Spot")
elif classes[0][5]==1:
print("Spider Mites")
elif classes[0][6]==1:
print("Target Spot")
elif classes[0][7]==1:
print("Tomato Yellow Leaf Curl Virus")
elif classes[0][8]==1:
print("Tomato Mosaic Virus")
else:
print("Healthy")
Saving healthy.jpg to healthy.jpg 1/1 [==============================] - 1s 782ms/step 1 Early blight
Klasifikasi PEDANTO RESNET152V2¶
Modules & Packages¶
In [1]:
# Basic Imports
import numpy as np
import matplotlib.pyplot as plt
from glob import glob
from PIL import Image
# Imports untuk membangun model
from tensorflow import keras
from keras.models import Model
from tensorflow.keras.applications import ResNet152V2
from tensorflow.keras.layers import GlobalAveragePooling2D, Dense
from keras.preprocessing import image
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential, load_model
# Ignore Warnings
import warnings
warnings.filterwarnings("ignore")
Import Dataset¶
In [2]:
from google.colab import drive
drive.mount("/content/gdrive")
Mounted at /content/gdrive
In [3]:
import os
os.chdir('/content/gdrive/MyDrive/Colab Notebooks/dataset8020/')
os.listdir()
Out[3]:
['vgg16_model.h5', 'model_inception.h5', 'model_Resnet152V2.h5', 'train', 'val', 'model_resnet152V2.h5']
In [4]:
# Folders in the Train Set
folders = glob('/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/*')
folders
Out[4]:
['/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Bacterial_spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Early_blight', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Healthy', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Late_blight', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Leaf_Mold', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Mosaic_virus', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Septoria_leaf_spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Spider_mites', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Target_Spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/train/Tomato___Yellow_Leaf_Curl_Virus']
In [5]:
# Folders in the Test Set
folders = glob('/content/gdrive/My Drive/Colab Notebooks/dataset8020/val/*')
folders
Out[5]:
['/content/gdrive/My Drive/Colab Notebooks/dataset8020/val/Tomato___Bacterial_spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/val/Tomato___Early_blight', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/val/Tomato___Healthy', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/val/Tomato___Late_blight', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/val/Tomato___Leaf_Mold', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/val/Tomato___Mosaic_virus', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/val/Tomato___Septoria_leaf_spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/val/Tomato___Spider_mites', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/val/Tomato___Target_Spot', '/content/gdrive/My Drive/Colab Notebooks/dataset8020/val/Tomato___Yellow_Leaf_Curl_Virus']
In [6]:
#Forming Dataset
datagen = keras.preprocessing.image.ImageDataGenerator(rescale=1/255)
# Training and validation dataset
train = datagen.flow_from_directory('/content/gdrive/MyDrive/Colab Notebooks/dataset8020/train/', seed=123, subset='training')
# Test dataset for evaluation
datagen2 = keras.preprocessing.image.ImageDataGenerator(rescale=1/255)
test = datagen2.flow_from_directory('/content/gdrive/MyDrive/Colab Notebooks/dataset8020/val/')
Found 8000 images belonging to 10 classes. Found 2000 images belonging to 10 classes.
Exploratory Data Analysis¶
In [7]:
# Training data visualization
classes = os.listdir('/content/gdrive/MyDrive/Colab Notebooks/dataset8020/train')
plt.figure(figsize=(25,10))
for i in enumerate(classes):
pic = os.listdir('/content/gdrive/MyDrive/Colab Notebooks/dataset8020/train/'+i[1])[0]
image = Image.open('/content/gdrive/MyDrive/Colab Notebooks/dataset8020/train/'+i[1]+'/'+pic)
image = np.asarray(image)
plt.subplot(2,5,i[0]+1)
plt.title('{0} / Shape = {1}'.format(i[1], image.shape))
plt.imshow(image)
plt.show()
Data Preprocessing¶
In [8]:
def normalize(input_image, input_mask):
input_image = tf.cast(input_image, tf.float32) / 255.0
input_mask -= 1
return input_image, input_mask
Model RESNET152V2¶
In [9]:
# Function to initialize model (ResNet152V2)
def get_model():
base_model = ResNet152V2(input_shape=(256,256,3), include_top=False)
for layers in base_model.layers[:140]:
layers.trainable = False
for layers in base_model.layers[140:]:
layers.trainable = True
x = base_model.output
x = GlobalAveragePooling2D()(x)
x = Dense(1000, activation='relu')(x)
pred = Dense(10, activation='softmax')(x)
model = Model(inputs=base_model.input, outputs=pred)
return model
In [10]:
# Model initialization
model = get_model()
model.summary()
Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/resnet/resnet152v2_weights_tf_dim_ordering_tf_kernels_notop.h5
234545216/234545216 [==============================] - 11s 0us/step
Model: "model"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_1 (InputLayer) [(None, 256, 256, 3 0 []
)]
conv1_pad (ZeroPadding2D) (None, 262, 262, 3) 0 ['input_1[0][0]']
conv1_conv (Conv2D) (None, 128, 128, 64 9472 ['conv1_pad[0][0]']
)
pool1_pad (ZeroPadding2D) (None, 130, 130, 64 0 ['conv1_conv[0][0]']
)
pool1_pool (MaxPooling2D) (None, 64, 64, 64) 0 ['pool1_pad[0][0]']
conv2_block1_preact_bn (BatchN (None, 64, 64, 64) 256 ['pool1_pool[0][0]']
ormalization)
conv2_block1_preact_relu (Acti (None, 64, 64, 64) 0 ['conv2_block1_preact_bn[0][0]']
vation)
conv2_block1_1_conv (Conv2D) (None, 64, 64, 64) 4096 ['conv2_block1_preact_relu[0][0]'
]
conv2_block1_1_bn (BatchNormal (None, 64, 64, 64) 256 ['conv2_block1_1_conv[0][0]']
ization)
conv2_block1_1_relu (Activatio (None, 64, 64, 64) 0 ['conv2_block1_1_bn[0][0]']
n)
conv2_block1_2_pad (ZeroPaddin (None, 66, 66, 64) 0 ['conv2_block1_1_relu[0][0]']
g2D)
conv2_block1_2_conv (Conv2D) (None, 64, 64, 64) 36864 ['conv2_block1_2_pad[0][0]']
conv2_block1_2_bn (BatchNormal (None, 64, 64, 64) 256 ['conv2_block1_2_conv[0][0]']
ization)
conv2_block1_2_relu (Activatio (None, 64, 64, 64) 0 ['conv2_block1_2_bn[0][0]']
n)
conv2_block1_0_conv (Conv2D) (None, 64, 64, 256) 16640 ['conv2_block1_preact_relu[0][0]'
]
conv2_block1_3_conv (Conv2D) (None, 64, 64, 256) 16640 ['conv2_block1_2_relu[0][0]']
conv2_block1_out (Add) (None, 64, 64, 256) 0 ['conv2_block1_0_conv[0][0]',
'conv2_block1_3_conv[0][0]']
conv2_block2_preact_bn (BatchN (None, 64, 64, 256) 1024 ['conv2_block1_out[0][0]']
ormalization)
conv2_block2_preact_relu (Acti (None, 64, 64, 256) 0 ['conv2_block2_preact_bn[0][0]']
vation)
conv2_block2_1_conv (Conv2D) (None, 64, 64, 64) 16384 ['conv2_block2_preact_relu[0][0]'
]
conv2_block2_1_bn (BatchNormal (None, 64, 64, 64) 256 ['conv2_block2_1_conv[0][0]']
ization)
conv2_block2_1_relu (Activatio (None, 64, 64, 64) 0 ['conv2_block2_1_bn[0][0]']
n)
conv2_block2_2_pad (ZeroPaddin (None, 66, 66, 64) 0 ['conv2_block2_1_relu[0][0]']
g2D)
conv2_block2_2_conv (Conv2D) (None, 64, 64, 64) 36864 ['conv2_block2_2_pad[0][0]']
conv2_block2_2_bn (BatchNormal (None, 64, 64, 64) 256 ['conv2_block2_2_conv[0][0]']
ization)
conv2_block2_2_relu (Activatio (None, 64, 64, 64) 0 ['conv2_block2_2_bn[0][0]']
n)
conv2_block2_3_conv (Conv2D) (None, 64, 64, 256) 16640 ['conv2_block2_2_relu[0][0]']
conv2_block2_out (Add) (None, 64, 64, 256) 0 ['conv2_block1_out[0][0]',
'conv2_block2_3_conv[0][0]']
conv2_block3_preact_bn (BatchN (None, 64, 64, 256) 1024 ['conv2_block2_out[0][0]']
ormalization)
conv2_block3_preact_relu (Acti (None, 64, 64, 256) 0 ['conv2_block3_preact_bn[0][0]']
vation)
conv2_block3_1_conv (Conv2D) (None, 64, 64, 64) 16384 ['conv2_block3_preact_relu[0][0]'
]
conv2_block3_1_bn (BatchNormal (None, 64, 64, 64) 256 ['conv2_block3_1_conv[0][0]']
ization)
conv2_block3_1_relu (Activatio (None, 64, 64, 64) 0 ['conv2_block3_1_bn[0][0]']
n)
conv2_block3_2_pad (ZeroPaddin (None, 66, 66, 64) 0 ['conv2_block3_1_relu[0][0]']
g2D)
conv2_block3_2_conv (Conv2D) (None, 32, 32, 64) 36864 ['conv2_block3_2_pad[0][0]']
conv2_block3_2_bn (BatchNormal (None, 32, 32, 64) 256 ['conv2_block3_2_conv[0][0]']
ization)
conv2_block3_2_relu (Activatio (None, 32, 32, 64) 0 ['conv2_block3_2_bn[0][0]']
n)
max_pooling2d (MaxPooling2D) (None, 32, 32, 256) 0 ['conv2_block2_out[0][0]']
conv2_block3_3_conv (Conv2D) (None, 32, 32, 256) 16640 ['conv2_block3_2_relu[0][0]']
conv2_block3_out (Add) (None, 32, 32, 256) 0 ['max_pooling2d[0][0]',
'conv2_block3_3_conv[0][0]']
conv3_block1_preact_bn (BatchN (None, 32, 32, 256) 1024 ['conv2_block3_out[0][0]']
ormalization)
conv3_block1_preact_relu (Acti (None, 32, 32, 256) 0 ['conv3_block1_preact_bn[0][0]']
vation)
conv3_block1_1_conv (Conv2D) (None, 32, 32, 128) 32768 ['conv3_block1_preact_relu[0][0]'
]
conv3_block1_1_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block1_1_conv[0][0]']
ization)
conv3_block1_1_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block1_1_bn[0][0]']
n)
conv3_block1_2_pad (ZeroPaddin (None, 34, 34, 128) 0 ['conv3_block1_1_relu[0][0]']
g2D)
conv3_block1_2_conv (Conv2D) (None, 32, 32, 128) 147456 ['conv3_block1_2_pad[0][0]']
conv3_block1_2_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block1_2_conv[0][0]']
ization)
conv3_block1_2_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block1_2_bn[0][0]']
n)
conv3_block1_0_conv (Conv2D) (None, 32, 32, 512) 131584 ['conv3_block1_preact_relu[0][0]'
]
conv3_block1_3_conv (Conv2D) (None, 32, 32, 512) 66048 ['conv3_block1_2_relu[0][0]']
conv3_block1_out (Add) (None, 32, 32, 512) 0 ['conv3_block1_0_conv[0][0]',
'conv3_block1_3_conv[0][0]']
conv3_block2_preact_bn (BatchN (None, 32, 32, 512) 2048 ['conv3_block1_out[0][0]']
ormalization)
conv3_block2_preact_relu (Acti (None, 32, 32, 512) 0 ['conv3_block2_preact_bn[0][0]']
vation)
conv3_block2_1_conv (Conv2D) (None, 32, 32, 128) 65536 ['conv3_block2_preact_relu[0][0]'
]
conv3_block2_1_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block2_1_conv[0][0]']
ization)
conv3_block2_1_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block2_1_bn[0][0]']
n)
conv3_block2_2_pad (ZeroPaddin (None, 34, 34, 128) 0 ['conv3_block2_1_relu[0][0]']
g2D)
conv3_block2_2_conv (Conv2D) (None, 32, 32, 128) 147456 ['conv3_block2_2_pad[0][0]']
conv3_block2_2_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block2_2_conv[0][0]']
ization)
conv3_block2_2_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block2_2_bn[0][0]']
n)
conv3_block2_3_conv (Conv2D) (None, 32, 32, 512) 66048 ['conv3_block2_2_relu[0][0]']
conv3_block2_out (Add) (None, 32, 32, 512) 0 ['conv3_block1_out[0][0]',
'conv3_block2_3_conv[0][0]']
conv3_block3_preact_bn (BatchN (None, 32, 32, 512) 2048 ['conv3_block2_out[0][0]']
ormalization)
conv3_block3_preact_relu (Acti (None, 32, 32, 512) 0 ['conv3_block3_preact_bn[0][0]']
vation)
conv3_block3_1_conv (Conv2D) (None, 32, 32, 128) 65536 ['conv3_block3_preact_relu[0][0]'
]
conv3_block3_1_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block3_1_conv[0][0]']
ization)
conv3_block3_1_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block3_1_bn[0][0]']
n)
conv3_block3_2_pad (ZeroPaddin (None, 34, 34, 128) 0 ['conv3_block3_1_relu[0][0]']
g2D)
conv3_block3_2_conv (Conv2D) (None, 32, 32, 128) 147456 ['conv3_block3_2_pad[0][0]']
conv3_block3_2_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block3_2_conv[0][0]']
ization)
conv3_block3_2_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block3_2_bn[0][0]']
n)
conv3_block3_3_conv (Conv2D) (None, 32, 32, 512) 66048 ['conv3_block3_2_relu[0][0]']
conv3_block3_out (Add) (None, 32, 32, 512) 0 ['conv3_block2_out[0][0]',
'conv3_block3_3_conv[0][0]']
conv3_block4_preact_bn (BatchN (None, 32, 32, 512) 2048 ['conv3_block3_out[0][0]']
ormalization)
conv3_block4_preact_relu (Acti (None, 32, 32, 512) 0 ['conv3_block4_preact_bn[0][0]']
vation)
conv3_block4_1_conv (Conv2D) (None, 32, 32, 128) 65536 ['conv3_block4_preact_relu[0][0]'
]
conv3_block4_1_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block4_1_conv[0][0]']
ization)
conv3_block4_1_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block4_1_bn[0][0]']
n)
conv3_block4_2_pad (ZeroPaddin (None, 34, 34, 128) 0 ['conv3_block4_1_relu[0][0]']
g2D)
conv3_block4_2_conv (Conv2D) (None, 32, 32, 128) 147456 ['conv3_block4_2_pad[0][0]']
conv3_block4_2_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block4_2_conv[0][0]']
ization)
conv3_block4_2_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block4_2_bn[0][0]']
n)
conv3_block4_3_conv (Conv2D) (None, 32, 32, 512) 66048 ['conv3_block4_2_relu[0][0]']
conv3_block4_out (Add) (None, 32, 32, 512) 0 ['conv3_block3_out[0][0]',
'conv3_block4_3_conv[0][0]']
conv3_block5_preact_bn (BatchN (None, 32, 32, 512) 2048 ['conv3_block4_out[0][0]']
ormalization)
conv3_block5_preact_relu (Acti (None, 32, 32, 512) 0 ['conv3_block5_preact_bn[0][0]']
vation)
conv3_block5_1_conv (Conv2D) (None, 32, 32, 128) 65536 ['conv3_block5_preact_relu[0][0]'
]
conv3_block5_1_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block5_1_conv[0][0]']
ization)
conv3_block5_1_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block5_1_bn[0][0]']
n)
conv3_block5_2_pad (ZeroPaddin (None, 34, 34, 128) 0 ['conv3_block5_1_relu[0][0]']
g2D)
conv3_block5_2_conv (Conv2D) (None, 32, 32, 128) 147456 ['conv3_block5_2_pad[0][0]']
conv3_block5_2_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block5_2_conv[0][0]']
ization)
conv3_block5_2_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block5_2_bn[0][0]']
n)
conv3_block5_3_conv (Conv2D) (None, 32, 32, 512) 66048 ['conv3_block5_2_relu[0][0]']
conv3_block5_out (Add) (None, 32, 32, 512) 0 ['conv3_block4_out[0][0]',
'conv3_block5_3_conv[0][0]']
conv3_block6_preact_bn (BatchN (None, 32, 32, 512) 2048 ['conv3_block5_out[0][0]']
ormalization)
conv3_block6_preact_relu (Acti (None, 32, 32, 512) 0 ['conv3_block6_preact_bn[0][0]']
vation)
conv3_block6_1_conv (Conv2D) (None, 32, 32, 128) 65536 ['conv3_block6_preact_relu[0][0]'
]
conv3_block6_1_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block6_1_conv[0][0]']
ization)
conv3_block6_1_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block6_1_bn[0][0]']
n)
conv3_block6_2_pad (ZeroPaddin (None, 34, 34, 128) 0 ['conv3_block6_1_relu[0][0]']
g2D)
conv3_block6_2_conv (Conv2D) (None, 32, 32, 128) 147456 ['conv3_block6_2_pad[0][0]']
conv3_block6_2_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block6_2_conv[0][0]']
ization)
conv3_block6_2_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block6_2_bn[0][0]']
n)
conv3_block6_3_conv (Conv2D) (None, 32, 32, 512) 66048 ['conv3_block6_2_relu[0][0]']
conv3_block6_out (Add) (None, 32, 32, 512) 0 ['conv3_block5_out[0][0]',
'conv3_block6_3_conv[0][0]']
conv3_block7_preact_bn (BatchN (None, 32, 32, 512) 2048 ['conv3_block6_out[0][0]']
ormalization)
conv3_block7_preact_relu (Acti (None, 32, 32, 512) 0 ['conv3_block7_preact_bn[0][0]']
vation)
conv3_block7_1_conv (Conv2D) (None, 32, 32, 128) 65536 ['conv3_block7_preact_relu[0][0]'
]
conv3_block7_1_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block7_1_conv[0][0]']
ization)
conv3_block7_1_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block7_1_bn[0][0]']
n)
conv3_block7_2_pad (ZeroPaddin (None, 34, 34, 128) 0 ['conv3_block7_1_relu[0][0]']
g2D)
conv3_block7_2_conv (Conv2D) (None, 32, 32, 128) 147456 ['conv3_block7_2_pad[0][0]']
conv3_block7_2_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block7_2_conv[0][0]']
ization)
conv3_block7_2_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block7_2_bn[0][0]']
n)
conv3_block7_3_conv (Conv2D) (None, 32, 32, 512) 66048 ['conv3_block7_2_relu[0][0]']
conv3_block7_out (Add) (None, 32, 32, 512) 0 ['conv3_block6_out[0][0]',
'conv3_block7_3_conv[0][0]']
conv3_block8_preact_bn (BatchN (None, 32, 32, 512) 2048 ['conv3_block7_out[0][0]']
ormalization)
conv3_block8_preact_relu (Acti (None, 32, 32, 512) 0 ['conv3_block8_preact_bn[0][0]']
vation)
conv3_block8_1_conv (Conv2D) (None, 32, 32, 128) 65536 ['conv3_block8_preact_relu[0][0]'
]
conv3_block8_1_bn (BatchNormal (None, 32, 32, 128) 512 ['conv3_block8_1_conv[0][0]']
ization)
conv3_block8_1_relu (Activatio (None, 32, 32, 128) 0 ['conv3_block8_1_bn[0][0]']
n)
conv3_block8_2_pad (ZeroPaddin (None, 34, 34, 128) 0 ['conv3_block8_1_relu[0][0]']
g2D)
conv3_block8_2_conv (Conv2D) (None, 16, 16, 128) 147456 ['conv3_block8_2_pad[0][0]']
conv3_block8_2_bn (BatchNormal (None, 16, 16, 128) 512 ['conv3_block8_2_conv[0][0]']
ization)
conv3_block8_2_relu (Activatio (None, 16, 16, 128) 0 ['conv3_block8_2_bn[0][0]']
n)
max_pooling2d_1 (MaxPooling2D) (None, 16, 16, 512) 0 ['conv3_block7_out[0][0]']
conv3_block8_3_conv (Conv2D) (None, 16, 16, 512) 66048 ['conv3_block8_2_relu[0][0]']
conv3_block8_out (Add) (None, 16, 16, 512) 0 ['max_pooling2d_1[0][0]',
'conv3_block8_3_conv[0][0]']
conv4_block1_preact_bn (BatchN (None, 16, 16, 512) 2048 ['conv3_block8_out[0][0]']
ormalization)
conv4_block1_preact_relu (Acti (None, 16, 16, 512) 0 ['conv4_block1_preact_bn[0][0]']
vation)
conv4_block1_1_conv (Conv2D) (None, 16, 16, 256) 131072 ['conv4_block1_preact_relu[0][0]'
]
conv4_block1_1_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block1_1_conv[0][0]']
ization)
conv4_block1_1_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block1_1_bn[0][0]']
n)
conv4_block1_2_pad (ZeroPaddin (None, 18, 18, 256) 0 ['conv4_block1_1_relu[0][0]']
g2D)
conv4_block1_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block1_2_pad[0][0]']
conv4_block1_2_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block1_2_conv[0][0]']
ization)
conv4_block1_2_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block1_2_bn[0][0]']
n)
conv4_block1_0_conv (Conv2D) (None, 16, 16, 1024 525312 ['conv4_block1_preact_relu[0][0]'
) ]
conv4_block1_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block1_2_relu[0][0]']
)
conv4_block1_out (Add) (None, 16, 16, 1024 0 ['conv4_block1_0_conv[0][0]',
) 'conv4_block1_3_conv[0][0]']
conv4_block2_preact_bn (BatchN (None, 16, 16, 1024 4096 ['conv4_block1_out[0][0]']
ormalization) )
conv4_block2_preact_relu (Acti (None, 16, 16, 1024 0 ['conv4_block2_preact_bn[0][0]']
vation) )
conv4_block2_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block2_preact_relu[0][0]'
]
conv4_block2_1_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block2_1_conv[0][0]']
ization)
conv4_block2_1_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block2_1_bn[0][0]']
n)
conv4_block2_2_pad (ZeroPaddin (None, 18, 18, 256) 0 ['conv4_block2_1_relu[0][0]']
g2D)
conv4_block2_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block2_2_pad[0][0]']
conv4_block2_2_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block2_2_conv[0][0]']
ization)
conv4_block2_2_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block2_2_bn[0][0]']
n)
conv4_block2_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block2_2_relu[0][0]']
)
conv4_block2_out (Add) (None, 16, 16, 1024 0 ['conv4_block1_out[0][0]',
) 'conv4_block2_3_conv[0][0]']
conv4_block3_preact_bn (BatchN (None, 16, 16, 1024 4096 ['conv4_block2_out[0][0]']
ormalization) )
conv4_block3_preact_relu (Acti (None, 16, 16, 1024 0 ['conv4_block3_preact_bn[0][0]']
vation) )
conv4_block3_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block3_preact_relu[0][0]'
]
conv4_block3_1_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block3_1_conv[0][0]']
ization)
conv4_block3_1_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block3_1_bn[0][0]']
n)
conv4_block3_2_pad (ZeroPaddin (None, 18, 18, 256) 0 ['conv4_block3_1_relu[0][0]']
g2D)
conv4_block3_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block3_2_pad[0][0]']
conv4_block3_2_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block3_2_conv[0][0]']
ization)
conv4_block3_2_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block3_2_bn[0][0]']
n)
conv4_block3_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block3_2_relu[0][0]']
)
conv4_block3_out (Add) (None, 16, 16, 1024 0 ['conv4_block2_out[0][0]',
) 'conv4_block3_3_conv[0][0]']
conv4_block4_preact_bn (BatchN (None, 16, 16, 1024 4096 ['conv4_block3_out[0][0]']
ormalization) )
conv4_block4_preact_relu (Acti (None, 16, 16, 1024 0 ['conv4_block4_preact_bn[0][0]']
vation) )
conv4_block4_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block4_preact_relu[0][0]'
]
conv4_block4_1_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block4_1_conv[0][0]']
ization)
conv4_block4_1_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block4_1_bn[0][0]']
n)
conv4_block4_2_pad (ZeroPaddin (None, 18, 18, 256) 0 ['conv4_block4_1_relu[0][0]']
g2D)
conv4_block4_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block4_2_pad[0][0]']
conv4_block4_2_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block4_2_conv[0][0]']
ization)
conv4_block4_2_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block4_2_bn[0][0]']
n)
conv4_block4_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block4_2_relu[0][0]']
)
conv4_block4_out (Add) (None, 16, 16, 1024 0 ['conv4_block3_out[0][0]',
) 'conv4_block4_3_conv[0][0]']
conv4_block5_preact_bn (BatchN (None, 16, 16, 1024 4096 ['conv4_block4_out[0][0]']
ormalization) )
conv4_block5_preact_relu (Acti (None, 16, 16, 1024 0 ['conv4_block5_preact_bn[0][0]']
vation) )
conv4_block5_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block5_preact_relu[0][0]'
]
conv4_block5_1_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block5_1_conv[0][0]']
ization)
conv4_block5_1_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block5_1_bn[0][0]']
n)
conv4_block5_2_pad (ZeroPaddin (None, 18, 18, 256) 0 ['conv4_block5_1_relu[0][0]']
g2D)
conv4_block5_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block5_2_pad[0][0]']
conv4_block5_2_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block5_2_conv[0][0]']
ization)
conv4_block5_2_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block5_2_bn[0][0]']
n)
conv4_block5_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block5_2_relu[0][0]']
)
conv4_block5_out (Add) (None, 16, 16, 1024 0 ['conv4_block4_out[0][0]',
) 'conv4_block5_3_conv[0][0]']
conv4_block6_preact_bn (BatchN (None, 16, 16, 1024 4096 ['conv4_block5_out[0][0]']
ormalization) )
conv4_block6_preact_relu (Acti (None, 16, 16, 1024 0 ['conv4_block6_preact_bn[0][0]']
vation) )
conv4_block6_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block6_preact_relu[0][0]'
]
conv4_block6_1_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block6_1_conv[0][0]']
ization)
conv4_block6_1_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block6_1_bn[0][0]']
n)
conv4_block6_2_pad (ZeroPaddin (None, 18, 18, 256) 0 ['conv4_block6_1_relu[0][0]']
g2D)
conv4_block6_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block6_2_pad[0][0]']
conv4_block6_2_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block6_2_conv[0][0]']
ization)
conv4_block6_2_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block6_2_bn[0][0]']
n)
conv4_block6_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block6_2_relu[0][0]']
)
conv4_block6_out (Add) (None, 16, 16, 1024 0 ['conv4_block5_out[0][0]',
) 'conv4_block6_3_conv[0][0]']
conv4_block7_preact_bn (BatchN (None, 16, 16, 1024 4096 ['conv4_block6_out[0][0]']
ormalization) )
conv4_block7_preact_relu (Acti (None, 16, 16, 1024 0 ['conv4_block7_preact_bn[0][0]']
vation) )
conv4_block7_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block7_preact_relu[0][0]'
]
conv4_block7_1_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block7_1_conv[0][0]']
ization)
conv4_block7_1_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block7_1_bn[0][0]']
n)
conv4_block7_2_pad (ZeroPaddin (None, 18, 18, 256) 0 ['conv4_block7_1_relu[0][0]']
g2D)
conv4_block7_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block7_2_pad[0][0]']
conv4_block7_2_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block7_2_conv[0][0]']
ization)
conv4_block7_2_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block7_2_bn[0][0]']
n)
conv4_block7_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block7_2_relu[0][0]']
)
conv4_block7_out (Add) (None, 16, 16, 1024 0 ['conv4_block6_out[0][0]',
) 'conv4_block7_3_conv[0][0]']
conv4_block8_preact_bn (BatchN (None, 16, 16, 1024 4096 ['conv4_block7_out[0][0]']
ormalization) )
conv4_block8_preact_relu (Acti (None, 16, 16, 1024 0 ['conv4_block8_preact_bn[0][0]']
vation) )
conv4_block8_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block8_preact_relu[0][0]'
]
conv4_block8_1_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block8_1_conv[0][0]']
ization)
conv4_block8_1_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block8_1_bn[0][0]']
n)
conv4_block8_2_pad (ZeroPaddin (None, 18, 18, 256) 0 ['conv4_block8_1_relu[0][0]']
g2D)
conv4_block8_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block8_2_pad[0][0]']
conv4_block8_2_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block8_2_conv[0][0]']
ization)
conv4_block8_2_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block8_2_bn[0][0]']
n)
conv4_block8_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block8_2_relu[0][0]']
)
conv4_block8_out (Add) (None, 16, 16, 1024 0 ['conv4_block7_out[0][0]',
) 'conv4_block8_3_conv[0][0]']
conv4_block9_preact_bn (BatchN (None, 16, 16, 1024 4096 ['conv4_block8_out[0][0]']
ormalization) )
conv4_block9_preact_relu (Acti (None, 16, 16, 1024 0 ['conv4_block9_preact_bn[0][0]']
vation) )
conv4_block9_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block9_preact_relu[0][0]'
]
conv4_block9_1_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block9_1_conv[0][0]']
ization)
conv4_block9_1_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block9_1_bn[0][0]']
n)
conv4_block9_2_pad (ZeroPaddin (None, 18, 18, 256) 0 ['conv4_block9_1_relu[0][0]']
g2D)
conv4_block9_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block9_2_pad[0][0]']
conv4_block9_2_bn (BatchNormal (None, 16, 16, 256) 1024 ['conv4_block9_2_conv[0][0]']
ization)
conv4_block9_2_relu (Activatio (None, 16, 16, 256) 0 ['conv4_block9_2_bn[0][0]']
n)
conv4_block9_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block9_2_relu[0][0]']
)
conv4_block9_out (Add) (None, 16, 16, 1024 0 ['conv4_block8_out[0][0]',
) 'conv4_block9_3_conv[0][0]']
conv4_block10_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block9_out[0][0]']
Normalization) )
conv4_block10_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block10_preact_bn[0][0]']
ivation) )
conv4_block10_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block10_preact_relu[0][0]
']
conv4_block10_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block10_1_conv[0][0]']
lization)
conv4_block10_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block10_1_bn[0][0]']
on)
conv4_block10_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block10_1_relu[0][0]']
ng2D)
conv4_block10_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block10_2_pad[0][0]']
conv4_block10_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block10_2_conv[0][0]']
lization)
conv4_block10_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block10_2_bn[0][0]']
on)
conv4_block10_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block10_2_relu[0][0]']
)
conv4_block10_out (Add) (None, 16, 16, 1024 0 ['conv4_block9_out[0][0]',
) 'conv4_block10_3_conv[0][0]']
conv4_block11_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block10_out[0][0]']
Normalization) )
conv4_block11_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block11_preact_bn[0][0]']
ivation) )
conv4_block11_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block11_preact_relu[0][0]
']
conv4_block11_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block11_1_conv[0][0]']
lization)
conv4_block11_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block11_1_bn[0][0]']
on)
conv4_block11_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block11_1_relu[0][0]']
ng2D)
conv4_block11_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block11_2_pad[0][0]']
conv4_block11_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block11_2_conv[0][0]']
lization)
conv4_block11_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block11_2_bn[0][0]']
on)
conv4_block11_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block11_2_relu[0][0]']
)
conv4_block11_out (Add) (None, 16, 16, 1024 0 ['conv4_block10_out[0][0]',
) 'conv4_block11_3_conv[0][0]']
conv4_block12_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block11_out[0][0]']
Normalization) )
conv4_block12_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block12_preact_bn[0][0]']
ivation) )
conv4_block12_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block12_preact_relu[0][0]
']
conv4_block12_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block12_1_conv[0][0]']
lization)
conv4_block12_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block12_1_bn[0][0]']
on)
conv4_block12_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block12_1_relu[0][0]']
ng2D)
conv4_block12_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block12_2_pad[0][0]']
conv4_block12_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block12_2_conv[0][0]']
lization)
conv4_block12_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block12_2_bn[0][0]']
on)
conv4_block12_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block12_2_relu[0][0]']
)
conv4_block12_out (Add) (None, 16, 16, 1024 0 ['conv4_block11_out[0][0]',
) 'conv4_block12_3_conv[0][0]']
conv4_block13_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block12_out[0][0]']
Normalization) )
conv4_block13_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block13_preact_bn[0][0]']
ivation) )
conv4_block13_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block13_preact_relu[0][0]
']
conv4_block13_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block13_1_conv[0][0]']
lization)
conv4_block13_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block13_1_bn[0][0]']
on)
conv4_block13_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block13_1_relu[0][0]']
ng2D)
conv4_block13_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block13_2_pad[0][0]']
conv4_block13_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block13_2_conv[0][0]']
lization)
conv4_block13_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block13_2_bn[0][0]']
on)
conv4_block13_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block13_2_relu[0][0]']
)
conv4_block13_out (Add) (None, 16, 16, 1024 0 ['conv4_block12_out[0][0]',
) 'conv4_block13_3_conv[0][0]']
conv4_block14_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block13_out[0][0]']
Normalization) )
conv4_block14_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block14_preact_bn[0][0]']
ivation) )
conv4_block14_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block14_preact_relu[0][0]
']
conv4_block14_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block14_1_conv[0][0]']
lization)
conv4_block14_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block14_1_bn[0][0]']
on)
conv4_block14_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block14_1_relu[0][0]']
ng2D)
conv4_block14_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block14_2_pad[0][0]']
conv4_block14_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block14_2_conv[0][0]']
lization)
conv4_block14_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block14_2_bn[0][0]']
on)
conv4_block14_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block14_2_relu[0][0]']
)
conv4_block14_out (Add) (None, 16, 16, 1024 0 ['conv4_block13_out[0][0]',
) 'conv4_block14_3_conv[0][0]']
conv4_block15_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block14_out[0][0]']
Normalization) )
conv4_block15_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block15_preact_bn[0][0]']
ivation) )
conv4_block15_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block15_preact_relu[0][0]
']
conv4_block15_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block15_1_conv[0][0]']
lization)
conv4_block15_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block15_1_bn[0][0]']
on)
conv4_block15_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block15_1_relu[0][0]']
ng2D)
conv4_block15_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block15_2_pad[0][0]']
conv4_block15_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block15_2_conv[0][0]']
lization)
conv4_block15_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block15_2_bn[0][0]']
on)
conv4_block15_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block15_2_relu[0][0]']
)
conv4_block15_out (Add) (None, 16, 16, 1024 0 ['conv4_block14_out[0][0]',
) 'conv4_block15_3_conv[0][0]']
conv4_block16_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block15_out[0][0]']
Normalization) )
conv4_block16_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block16_preact_bn[0][0]']
ivation) )
conv4_block16_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block16_preact_relu[0][0]
']
conv4_block16_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block16_1_conv[0][0]']
lization)
conv4_block16_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block16_1_bn[0][0]']
on)
conv4_block16_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block16_1_relu[0][0]']
ng2D)
conv4_block16_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block16_2_pad[0][0]']
conv4_block16_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block16_2_conv[0][0]']
lization)
conv4_block16_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block16_2_bn[0][0]']
on)
conv4_block16_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block16_2_relu[0][0]']
)
conv4_block16_out (Add) (None, 16, 16, 1024 0 ['conv4_block15_out[0][0]',
) 'conv4_block16_3_conv[0][0]']
conv4_block17_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block16_out[0][0]']
Normalization) )
conv4_block17_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block17_preact_bn[0][0]']
ivation) )
conv4_block17_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block17_preact_relu[0][0]
']
conv4_block17_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block17_1_conv[0][0]']
lization)
conv4_block17_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block17_1_bn[0][0]']
on)
conv4_block17_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block17_1_relu[0][0]']
ng2D)
conv4_block17_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block17_2_pad[0][0]']
conv4_block17_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block17_2_conv[0][0]']
lization)
conv4_block17_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block17_2_bn[0][0]']
on)
conv4_block17_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block17_2_relu[0][0]']
)
conv4_block17_out (Add) (None, 16, 16, 1024 0 ['conv4_block16_out[0][0]',
) 'conv4_block17_3_conv[0][0]']
conv4_block18_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block17_out[0][0]']
Normalization) )
conv4_block18_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block18_preact_bn[0][0]']
ivation) )
conv4_block18_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block18_preact_relu[0][0]
']
conv4_block18_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block18_1_conv[0][0]']
lization)
conv4_block18_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block18_1_bn[0][0]']
on)
conv4_block18_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block18_1_relu[0][0]']
ng2D)
conv4_block18_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block18_2_pad[0][0]']
conv4_block18_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block18_2_conv[0][0]']
lization)
conv4_block18_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block18_2_bn[0][0]']
on)
conv4_block18_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block18_2_relu[0][0]']
)
conv4_block18_out (Add) (None, 16, 16, 1024 0 ['conv4_block17_out[0][0]',
) 'conv4_block18_3_conv[0][0]']
conv4_block19_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block18_out[0][0]']
Normalization) )
conv4_block19_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block19_preact_bn[0][0]']
ivation) )
conv4_block19_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block19_preact_relu[0][0]
']
conv4_block19_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block19_1_conv[0][0]']
lization)
conv4_block19_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block19_1_bn[0][0]']
on)
conv4_block19_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block19_1_relu[0][0]']
ng2D)
conv4_block19_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block19_2_pad[0][0]']
conv4_block19_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block19_2_conv[0][0]']
lization)
conv4_block19_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block19_2_bn[0][0]']
on)
conv4_block19_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block19_2_relu[0][0]']
)
conv4_block19_out (Add) (None, 16, 16, 1024 0 ['conv4_block18_out[0][0]',
) 'conv4_block19_3_conv[0][0]']
conv4_block20_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block19_out[0][0]']
Normalization) )
conv4_block20_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block20_preact_bn[0][0]']
ivation) )
conv4_block20_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block20_preact_relu[0][0]
']
conv4_block20_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block20_1_conv[0][0]']
lization)
conv4_block20_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block20_1_bn[0][0]']
on)
conv4_block20_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block20_1_relu[0][0]']
ng2D)
conv4_block20_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block20_2_pad[0][0]']
conv4_block20_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block20_2_conv[0][0]']
lization)
conv4_block20_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block20_2_bn[0][0]']
on)
conv4_block20_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block20_2_relu[0][0]']
)
conv4_block20_out (Add) (None, 16, 16, 1024 0 ['conv4_block19_out[0][0]',
) 'conv4_block20_3_conv[0][0]']
conv4_block21_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block20_out[0][0]']
Normalization) )
conv4_block21_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block21_preact_bn[0][0]']
ivation) )
conv4_block21_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block21_preact_relu[0][0]
']
conv4_block21_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block21_1_conv[0][0]']
lization)
conv4_block21_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block21_1_bn[0][0]']
on)
conv4_block21_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block21_1_relu[0][0]']
ng2D)
conv4_block21_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block21_2_pad[0][0]']
conv4_block21_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block21_2_conv[0][0]']
lization)
conv4_block21_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block21_2_bn[0][0]']
on)
conv4_block21_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block21_2_relu[0][0]']
)
conv4_block21_out (Add) (None, 16, 16, 1024 0 ['conv4_block20_out[0][0]',
) 'conv4_block21_3_conv[0][0]']
conv4_block22_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block21_out[0][0]']
Normalization) )
conv4_block22_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block22_preact_bn[0][0]']
ivation) )
conv4_block22_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block22_preact_relu[0][0]
']
conv4_block22_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block22_1_conv[0][0]']
lization)
conv4_block22_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block22_1_bn[0][0]']
on)
conv4_block22_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block22_1_relu[0][0]']
ng2D)
conv4_block22_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block22_2_pad[0][0]']
conv4_block22_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block22_2_conv[0][0]']
lization)
conv4_block22_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block22_2_bn[0][0]']
on)
conv4_block22_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block22_2_relu[0][0]']
)
conv4_block22_out (Add) (None, 16, 16, 1024 0 ['conv4_block21_out[0][0]',
) 'conv4_block22_3_conv[0][0]']
conv4_block23_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block22_out[0][0]']
Normalization) )
conv4_block23_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block23_preact_bn[0][0]']
ivation) )
conv4_block23_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block23_preact_relu[0][0]
']
conv4_block23_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block23_1_conv[0][0]']
lization)
conv4_block23_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block23_1_bn[0][0]']
on)
conv4_block23_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block23_1_relu[0][0]']
ng2D)
conv4_block23_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block23_2_pad[0][0]']
conv4_block23_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block23_2_conv[0][0]']
lization)
conv4_block23_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block23_2_bn[0][0]']
on)
conv4_block23_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block23_2_relu[0][0]']
)
conv4_block23_out (Add) (None, 16, 16, 1024 0 ['conv4_block22_out[0][0]',
) 'conv4_block23_3_conv[0][0]']
conv4_block24_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block23_out[0][0]']
Normalization) )
conv4_block24_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block24_preact_bn[0][0]']
ivation) )
conv4_block24_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block24_preact_relu[0][0]
']
conv4_block24_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block24_1_conv[0][0]']
lization)
conv4_block24_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block24_1_bn[0][0]']
on)
conv4_block24_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block24_1_relu[0][0]']
ng2D)
conv4_block24_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block24_2_pad[0][0]']
conv4_block24_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block24_2_conv[0][0]']
lization)
conv4_block24_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block24_2_bn[0][0]']
on)
conv4_block24_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block24_2_relu[0][0]']
)
conv4_block24_out (Add) (None, 16, 16, 1024 0 ['conv4_block23_out[0][0]',
) 'conv4_block24_3_conv[0][0]']
conv4_block25_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block24_out[0][0]']
Normalization) )
conv4_block25_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block25_preact_bn[0][0]']
ivation) )
conv4_block25_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block25_preact_relu[0][0]
']
conv4_block25_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block25_1_conv[0][0]']
lization)
conv4_block25_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block25_1_bn[0][0]']
on)
conv4_block25_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block25_1_relu[0][0]']
ng2D)
conv4_block25_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block25_2_pad[0][0]']
conv4_block25_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block25_2_conv[0][0]']
lization)
conv4_block25_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block25_2_bn[0][0]']
on)
conv4_block25_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block25_2_relu[0][0]']
)
conv4_block25_out (Add) (None, 16, 16, 1024 0 ['conv4_block24_out[0][0]',
) 'conv4_block25_3_conv[0][0]']
conv4_block26_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block25_out[0][0]']
Normalization) )
conv4_block26_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block26_preact_bn[0][0]']
ivation) )
conv4_block26_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block26_preact_relu[0][0]
']
conv4_block26_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block26_1_conv[0][0]']
lization)
conv4_block26_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block26_1_bn[0][0]']
on)
conv4_block26_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block26_1_relu[0][0]']
ng2D)
conv4_block26_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block26_2_pad[0][0]']
conv4_block26_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block26_2_conv[0][0]']
lization)
conv4_block26_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block26_2_bn[0][0]']
on)
conv4_block26_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block26_2_relu[0][0]']
)
conv4_block26_out (Add) (None, 16, 16, 1024 0 ['conv4_block25_out[0][0]',
) 'conv4_block26_3_conv[0][0]']
conv4_block27_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block26_out[0][0]']
Normalization) )
conv4_block27_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block27_preact_bn[0][0]']
ivation) )
conv4_block27_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block27_preact_relu[0][0]
']
conv4_block27_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block27_1_conv[0][0]']
lization)
conv4_block27_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block27_1_bn[0][0]']
on)
conv4_block27_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block27_1_relu[0][0]']
ng2D)
conv4_block27_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block27_2_pad[0][0]']
conv4_block27_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block27_2_conv[0][0]']
lization)
conv4_block27_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block27_2_bn[0][0]']
on)
conv4_block27_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block27_2_relu[0][0]']
)
conv4_block27_out (Add) (None, 16, 16, 1024 0 ['conv4_block26_out[0][0]',
) 'conv4_block27_3_conv[0][0]']
conv4_block28_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block27_out[0][0]']
Normalization) )
conv4_block28_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block28_preact_bn[0][0]']
ivation) )
conv4_block28_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block28_preact_relu[0][0]
']
conv4_block28_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block28_1_conv[0][0]']
lization)
conv4_block28_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block28_1_bn[0][0]']
on)
conv4_block28_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block28_1_relu[0][0]']
ng2D)
conv4_block28_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block28_2_pad[0][0]']
conv4_block28_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block28_2_conv[0][0]']
lization)
conv4_block28_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block28_2_bn[0][0]']
on)
conv4_block28_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block28_2_relu[0][0]']
)
conv4_block28_out (Add) (None, 16, 16, 1024 0 ['conv4_block27_out[0][0]',
) 'conv4_block28_3_conv[0][0]']
conv4_block29_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block28_out[0][0]']
Normalization) )
conv4_block29_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block29_preact_bn[0][0]']
ivation) )
conv4_block29_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block29_preact_relu[0][0]
']
conv4_block29_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block29_1_conv[0][0]']
lization)
conv4_block29_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block29_1_bn[0][0]']
on)
conv4_block29_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block29_1_relu[0][0]']
ng2D)
conv4_block29_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block29_2_pad[0][0]']
conv4_block29_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block29_2_conv[0][0]']
lization)
conv4_block29_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block29_2_bn[0][0]']
on)
conv4_block29_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block29_2_relu[0][0]']
)
conv4_block29_out (Add) (None, 16, 16, 1024 0 ['conv4_block28_out[0][0]',
) 'conv4_block29_3_conv[0][0]']
conv4_block30_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block29_out[0][0]']
Normalization) )
conv4_block30_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block30_preact_bn[0][0]']
ivation) )
conv4_block30_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block30_preact_relu[0][0]
']
conv4_block30_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block30_1_conv[0][0]']
lization)
conv4_block30_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block30_1_bn[0][0]']
on)
conv4_block30_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block30_1_relu[0][0]']
ng2D)
conv4_block30_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block30_2_pad[0][0]']
conv4_block30_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block30_2_conv[0][0]']
lization)
conv4_block30_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block30_2_bn[0][0]']
on)
conv4_block30_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block30_2_relu[0][0]']
)
conv4_block30_out (Add) (None, 16, 16, 1024 0 ['conv4_block29_out[0][0]',
) 'conv4_block30_3_conv[0][0]']
conv4_block31_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block30_out[0][0]']
Normalization) )
conv4_block31_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block31_preact_bn[0][0]']
ivation) )
conv4_block31_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block31_preact_relu[0][0]
']
conv4_block31_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block31_1_conv[0][0]']
lization)
conv4_block31_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block31_1_bn[0][0]']
on)
conv4_block31_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block31_1_relu[0][0]']
ng2D)
conv4_block31_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block31_2_pad[0][0]']
conv4_block31_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block31_2_conv[0][0]']
lization)
conv4_block31_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block31_2_bn[0][0]']
on)
conv4_block31_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block31_2_relu[0][0]']
)
conv4_block31_out (Add) (None, 16, 16, 1024 0 ['conv4_block30_out[0][0]',
) 'conv4_block31_3_conv[0][0]']
conv4_block32_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block31_out[0][0]']
Normalization) )
conv4_block32_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block32_preact_bn[0][0]']
ivation) )
conv4_block32_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block32_preact_relu[0][0]
']
conv4_block32_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block32_1_conv[0][0]']
lization)
conv4_block32_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block32_1_bn[0][0]']
on)
conv4_block32_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block32_1_relu[0][0]']
ng2D)
conv4_block32_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block32_2_pad[0][0]']
conv4_block32_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block32_2_conv[0][0]']
lization)
conv4_block32_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block32_2_bn[0][0]']
on)
conv4_block32_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block32_2_relu[0][0]']
)
conv4_block32_out (Add) (None, 16, 16, 1024 0 ['conv4_block31_out[0][0]',
) 'conv4_block32_3_conv[0][0]']
conv4_block33_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block32_out[0][0]']
Normalization) )
conv4_block33_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block33_preact_bn[0][0]']
ivation) )
conv4_block33_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block33_preact_relu[0][0]
']
conv4_block33_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block33_1_conv[0][0]']
lization)
conv4_block33_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block33_1_bn[0][0]']
on)
conv4_block33_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block33_1_relu[0][0]']
ng2D)
conv4_block33_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block33_2_pad[0][0]']
conv4_block33_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block33_2_conv[0][0]']
lization)
conv4_block33_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block33_2_bn[0][0]']
on)
conv4_block33_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block33_2_relu[0][0]']
)
conv4_block33_out (Add) (None, 16, 16, 1024 0 ['conv4_block32_out[0][0]',
) 'conv4_block33_3_conv[0][0]']
conv4_block34_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block33_out[0][0]']
Normalization) )
conv4_block34_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block34_preact_bn[0][0]']
ivation) )
conv4_block34_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block34_preact_relu[0][0]
']
conv4_block34_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block34_1_conv[0][0]']
lization)
conv4_block34_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block34_1_bn[0][0]']
on)
conv4_block34_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block34_1_relu[0][0]']
ng2D)
conv4_block34_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block34_2_pad[0][0]']
conv4_block34_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block34_2_conv[0][0]']
lization)
conv4_block34_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block34_2_bn[0][0]']
on)
conv4_block34_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block34_2_relu[0][0]']
)
conv4_block34_out (Add) (None, 16, 16, 1024 0 ['conv4_block33_out[0][0]',
) 'conv4_block34_3_conv[0][0]']
conv4_block35_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block34_out[0][0]']
Normalization) )
conv4_block35_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block35_preact_bn[0][0]']
ivation) )
conv4_block35_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block35_preact_relu[0][0]
']
conv4_block35_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block35_1_conv[0][0]']
lization)
conv4_block35_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block35_1_bn[0][0]']
on)
conv4_block35_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block35_1_relu[0][0]']
ng2D)
conv4_block35_2_conv (Conv2D) (None, 16, 16, 256) 589824 ['conv4_block35_2_pad[0][0]']
conv4_block35_2_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block35_2_conv[0][0]']
lization)
conv4_block35_2_relu (Activati (None, 16, 16, 256) 0 ['conv4_block35_2_bn[0][0]']
on)
conv4_block35_3_conv (Conv2D) (None, 16, 16, 1024 263168 ['conv4_block35_2_relu[0][0]']
)
conv4_block35_out (Add) (None, 16, 16, 1024 0 ['conv4_block34_out[0][0]',
) 'conv4_block35_3_conv[0][0]']
conv4_block36_preact_bn (Batch (None, 16, 16, 1024 4096 ['conv4_block35_out[0][0]']
Normalization) )
conv4_block36_preact_relu (Act (None, 16, 16, 1024 0 ['conv4_block36_preact_bn[0][0]']
ivation) )
conv4_block36_1_conv (Conv2D) (None, 16, 16, 256) 262144 ['conv4_block36_preact_relu[0][0]
']
conv4_block36_1_bn (BatchNorma (None, 16, 16, 256) 1024 ['conv4_block36_1_conv[0][0]']
lization)
conv4_block36_1_relu (Activati (None, 16, 16, 256) 0 ['conv4_block36_1_bn[0][0]']
on)
conv4_block36_2_pad (ZeroPaddi (None, 18, 18, 256) 0 ['conv4_block36_1_relu[0][0]']
ng2D)
conv4_block36_2_conv (Conv2D) (None, 8, 8, 256) 589824 ['conv4_block36_2_pad[0][0]']
conv4_block36_2_bn (BatchNorma (None, 8, 8, 256) 1024 ['conv4_block36_2_conv[0][0]']
lization)
conv4_block36_2_relu (Activati (None, 8, 8, 256) 0 ['conv4_block36_2_bn[0][0]']
on)
max_pooling2d_2 (MaxPooling2D) (None, 8, 8, 1024) 0 ['conv4_block35_out[0][0]']
conv4_block36_3_conv (Conv2D) (None, 8, 8, 1024) 263168 ['conv4_block36_2_relu[0][0]']
conv4_block36_out (Add) (None, 8, 8, 1024) 0 ['max_pooling2d_2[0][0]',
'conv4_block36_3_conv[0][0]']
conv5_block1_preact_bn (BatchN (None, 8, 8, 1024) 4096 ['conv4_block36_out[0][0]']
ormalization)
conv5_block1_preact_relu (Acti (None, 8, 8, 1024) 0 ['conv5_block1_preact_bn[0][0]']
vation)
conv5_block1_1_conv (Conv2D) (None, 8, 8, 512) 524288 ['conv5_block1_preact_relu[0][0]'
]
conv5_block1_1_bn (BatchNormal (None, 8, 8, 512) 2048 ['conv5_block1_1_conv[0][0]']
ization)
conv5_block1_1_relu (Activatio (None, 8, 8, 512) 0 ['conv5_block1_1_bn[0][0]']
n)
conv5_block1_2_pad (ZeroPaddin (None, 10, 10, 512) 0 ['conv5_block1_1_relu[0][0]']
g2D)
conv5_block1_2_conv (Conv2D) (None, 8, 8, 512) 2359296 ['conv5_block1_2_pad[0][0]']
conv5_block1_2_bn (BatchNormal (None, 8, 8, 512) 2048 ['conv5_block1_2_conv[0][0]']
ization)
conv5_block1_2_relu (Activatio (None, 8, 8, 512) 0 ['conv5_block1_2_bn[0][0]']
n)
conv5_block1_0_conv (Conv2D) (None, 8, 8, 2048) 2099200 ['conv5_block1_preact_relu[0][0]'
]
conv5_block1_3_conv (Conv2D) (None, 8, 8, 2048) 1050624 ['conv5_block1_2_relu[0][0]']
conv5_block1_out (Add) (None, 8, 8, 2048) 0 ['conv5_block1_0_conv[0][0]',
'conv5_block1_3_conv[0][0]']
conv5_block2_preact_bn (BatchN (None, 8, 8, 2048) 8192 ['conv5_block1_out[0][0]']
ormalization)
conv5_block2_preact_relu (Acti (None, 8, 8, 2048) 0 ['conv5_block2_preact_bn[0][0]']
vation)
conv5_block2_1_conv (Conv2D) (None, 8, 8, 512) 1048576 ['conv5_block2_preact_relu[0][0]'
]
conv5_block2_1_bn (BatchNormal (None, 8, 8, 512) 2048 ['conv5_block2_1_conv[0][0]']
ization)
conv5_block2_1_relu (Activatio (None, 8, 8, 512) 0 ['conv5_block2_1_bn[0][0]']
n)
conv5_block2_2_pad (ZeroPaddin (None, 10, 10, 512) 0 ['conv5_block2_1_relu[0][0]']
g2D)
conv5_block2_2_conv (Conv2D) (None, 8, 8, 512) 2359296 ['conv5_block2_2_pad[0][0]']
conv5_block2_2_bn (BatchNormal (None, 8, 8, 512) 2048 ['conv5_block2_2_conv[0][0]']
ization)
conv5_block2_2_relu (Activatio (None, 8, 8, 512) 0 ['conv5_block2_2_bn[0][0]']
n)
conv5_block2_3_conv (Conv2D) (None, 8, 8, 2048) 1050624 ['conv5_block2_2_relu[0][0]']
conv5_block2_out (Add) (None, 8, 8, 2048) 0 ['conv5_block1_out[0][0]',
'conv5_block2_3_conv[0][0]']
conv5_block3_preact_bn (BatchN (None, 8, 8, 2048) 8192 ['conv5_block2_out[0][0]']
ormalization)
conv5_block3_preact_relu (Acti (None, 8, 8, 2048) 0 ['conv5_block3_preact_bn[0][0]']
vation)
conv5_block3_1_conv (Conv2D) (None, 8, 8, 512) 1048576 ['conv5_block3_preact_relu[0][0]'
]
conv5_block3_1_bn (BatchNormal (None, 8, 8, 512) 2048 ['conv5_block3_1_conv[0][0]']
ization)
conv5_block3_1_relu (Activatio (None, 8, 8, 512) 0 ['conv5_block3_1_bn[0][0]']
n)
conv5_block3_2_pad (ZeroPaddin (None, 10, 10, 512) 0 ['conv5_block3_1_relu[0][0]']
g2D)
conv5_block3_2_conv (Conv2D) (None, 8, 8, 512) 2359296 ['conv5_block3_2_pad[0][0]']
conv5_block3_2_bn (BatchNormal (None, 8, 8, 512) 2048 ['conv5_block3_2_conv[0][0]']
ization)
conv5_block3_2_relu (Activatio (None, 8, 8, 512) 0 ['conv5_block3_2_bn[0][0]']
n)
conv5_block3_3_conv (Conv2D) (None, 8, 8, 2048) 1050624 ['conv5_block3_2_relu[0][0]']
conv5_block3_out (Add) (None, 8, 8, 2048) 0 ['conv5_block2_out[0][0]',
'conv5_block3_3_conv[0][0]']
post_bn (BatchNormalization) (None, 8, 8, 2048) 8192 ['conv5_block3_out[0][0]']
post_relu (Activation) (None, 8, 8, 2048) 0 ['post_bn[0][0]']
global_average_pooling2d (Glob (None, 2048) 0 ['post_relu[0][0]']
alAveragePooling2D)
dense (Dense) (None, 1000) 2049000 ['global_average_pooling2d[0][0]'
]
dense_1 (Dense) (None, 10) 10010 ['dense[0][0]']
==================================================================================================
Total params: 60,390,658
Trainable params: 56,430,338
Non-trainable params: 3,960,320
__________________________________________________________________________________________________
Training Model¶
In [11]:
# Model training
model.compile(loss='categorical_crossentropy', optimizer='sgd', metrics='accuracy')
model.fit(train, batch_size=32, epochs=5, validation_data=test)
Epoch 1/5 250/250 [==============================] - 6429s 26s/step - loss: 0.5649 - accuracy: 0.8292 - val_loss: 0.2556 - val_accuracy: 0.9155 Epoch 2/5 250/250 [==============================] - 199s 796ms/step - loss: 0.0687 - accuracy: 0.9824 - val_loss: 0.1668 - val_accuracy: 0.9460 Epoch 3/5 250/250 [==============================] - 209s 836ms/step - loss: 0.0216 - accuracy: 0.9958 - val_loss: 0.1762 - val_accuracy: 0.9485 Epoch 4/5 250/250 [==============================] - 204s 815ms/step - loss: 0.0135 - accuracy: 0.9974 - val_loss: 0.1525 - val_accuracy: 0.9555 Epoch 5/5 250/250 [==============================] - 208s 830ms/step - loss: 0.0136 - accuracy: 0.9965 - val_loss: 0.1401 - val_accuracy: 0.9545
Out[11]:
<keras.callbacks.History at 0x7f7c4c8cbac0>
Evaluate Model¶
In [12]:
model.evaluate(test)
63/63 [==============================] - 19s 301ms/step - loss: 0.1401 - accuracy: 0.9545
Out[12]:
[0.1401241570711136, 0.9545000195503235]
In [13]:
loss_final, accuracy_final = model.evaluate(test)
63/63 [==============================] - 19s 301ms/step - loss: 0.1401 - accuracy: 0.9545
In [14]:
print("Final loss: {:.2f}".format(loss_final))
print("Final accuracy: {:.2f}".format(accuracy_final))
Final loss: 0.14 Final accuracy: 0.95
Prediction¶
In [15]:
# Prediction and visualizations
classes = os.listdir('/content/gdrive/MyDrive/Colab Notebooks/dataset8020/val')
plt.figure(figsize=(18,28))
for i in enumerate(classes):
pic = os.listdir('/content/gdrive/MyDrive/Colab Notebooks/dataset8020/val/'+i[1])
pic = pic[np.random.randint(len(pic)-1)]
image = Image.open('/content/gdrive/MyDrive/Colab Notebooks/dataset8020/val/'+i[1]+'/'+pic)
image = np.asarray(image)
pred = np.argmax(model.predict(image.reshape(-1,256,256,3)/255))
for j in list(enumerate(list(test.class_indices.keys()))):
if pred == j[0]:
prediction = j[1]
plt.subplot(5,2,i[0]+1)
plt.title('Actual: {0} / Predicted: {1}'.format(i[1], prediction))
plt.imshow(image)
plt.show()
1/1 [==============================] - 3s 3s/step 1/1 [==============================] - 0s 36ms/step 1/1 [==============================] - 0s 36ms/step 1/1 [==============================] - 0s 33ms/step 1/1 [==============================] - 0s 33ms/step 1/1 [==============================] - 0s 36ms/step 1/1 [==============================] - 0s 34ms/step 1/1 [==============================] - 0s 32ms/step 1/1 [==============================] - 0s 33ms/step 1/1 [==============================] - 0s 32ms/step
Predict test¶
In [16]:
y_pred = model.predict(test)
y_pred = np.argmax(y_pred, axis=1)
63/63 [==============================] - 20s 294ms/step
Save Model¶
In [17]:
# Saving the model as a h5 file
from tensorflow.keras.models import load_model
model.save('model_resnet152V2_tf.h5')
Load Model¶
In [18]:
model = load_model('model_resnet152V2_tf.h5')
Prediction Uploads Image¶
In [ ]:
%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf
from tensorflow.keras.preprocessing import image
from google.colab import files
uploaded = files.upload()
for fn in uploaded.keys():
# predicting images
path = '/content/gdrive/MyDrive/Colab Notebooks/uploads/' + fn
img = image.load_img(path, target_size=(256, 256, 3))
x = image.img_to_array(img)
plt.imshow(x/255.)
x = np.expand_dims(x, axis=0)
images = np.vstack([x])
classes = model.predict(images, batch_size=32)
index = tf.argmax(classes[0], axis=0)
print(classes[0])
if classes[0][0]==1:
print("Bacterial spot")
elif classes[0][1]==1:
print("Early blight")
elif classes[0][2]==1:
print("Late blight")
elif classes[0][3]==1:
print("Leaf Mold")
elif classes[0][4]==1:
print("Septoria Leaf Spot")
elif classes[0][5]==1:
print("Spider Mites")
elif classes[0][6]==1:
print("Target Spot")
elif classes[0][7]==1:
print("Tomato Yellow Leaf Curl Virus")
elif classes[0][8]==1:
print("Tomato Mosaic Virus")
else:
print("Healthy")
Saving Tomato Yellow Leaf Curl Virus.JPG to Tomato Yellow Leaf Curl Virus.JPG 1/1 [==============================] - 1s 657ms/step [0. 0. 0. 0. 0. 0. 0. 0. 0. 1.] Healthy
Prediction Uploads Image Fixing¶
In [ ]:
%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf
from tensorflow.keras.preprocessing import image
from google.colab import files
uploaded = files.upload()
for fn in uploaded.keys():
# predicting images
# path = lokasi filenya, fn itu lokasi file local di google colab
path = '/content/gdrive/MyDrive/Colab Notebooks/uploads/' + fn
img = image.load_img(path, target_size=(256, 256, 3))
x = image.img_to_array(img)
#xx=np.asarray(img)
#print ("X:",x)
#print ("XX:",xx)
plt.imshow(x/255.)
x = np.expand_dims(x, axis=0)
images = np.asarray(img)
#print ("images: ", images)
pred = np.argmax(model.predict(images.reshape(-1,256,256,3)/255))
print ("pred:",pred)
if pred==0:
print("Spider Mites")
elif pred==1:
print("Septoria Leaf Spot")
elif pred==2:
print("Late blight")
elif pred==3:
print("Mosaic Virus")
elif pred==4:
print("Leaf Mold")
elif pred==5:
print("Healthy")
elif pred==6:
print("Bacterial Spot")
elif pred==7:
print("Early Blight")
elif pred==8:
print("Target Spot")
else:
print("Yellow Leaf Curl Virus")
Saving bacterial spot 1.JPG to bacterial spot 1.JPG 1/1 [==============================] - 0s 35ms/step pred: 0 Spider Mites
Predict Uploads Image Fix¶
In [19]:
%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf
from tensorflow.keras.preprocessing import image
from google.colab import files
uploaded = files.upload()
for fn in uploaded.keys():
# predicting images
# path = lokasi filenya, fn itu lokasi file local di google colab
path = '/content/gdrive/MyDrive/Colab Notebooks/uploads/' + fn
img = image.load_img(path, target_size=(256, 256, 3))
x = image.img_to_array(img)
#xx=np.asarray(img)
#print ("X:",x)
#print ("XX:",xx)
plt.imshow(x/255.)
x = np.expand_dims(x, axis=0)
images = np.asarray(img)
#print ("images: ", images)
pred = np.argmax(model.predict(images.reshape(-1,256,256,3)/255))
print ("pred:",pred)
if pred==0:
print("Bacterial Spot")
elif pred==1:
print("Early Blight")
elif pred==2:
print("Healthy")
elif pred==3:
print("Late Blight")
elif pred==4:
print("Leaf Mold")
elif pred==5:
print("Mosaic Virus")
elif pred==6:
print("Septoria Leaf Spot")
elif pred==7:
print("Spider Mites")
elif pred==8:
print("Target Spot")
else:
print("Yellow Leaf Curl Virus")
Saving bacterial spot 1.JPG to bacterial spot 1.JPG 1/1 [==============================] - 2s 2s/step pred: 0 Bacterial Spot